The principle is simple: the controls stay fixed; the work product changes.

That boundary prevents an agent from making success easier after a weak attempt. If the specification, verifier, sources, tools, permissions, or decision rules need to change, the current run should stop. The agent may propose a change for a future run, but it should not approve its own new rules.

Define the Run Before It Starts

A real autonomous loop needs more than “keep improving.” It needs a retained baseline, a comparison rule, a budget, and explicit stopping conditions.

Andrej Karpathy’s autoresearch project demonstrates the pattern in a tightly controlled setting: an agent modifies one file, trains for a fixed five-minute budget, checks one objective metric, keeps or discards the result, records the experiment, and repeats [1]. The transferable lesson is not the specific code. It is the design of the environment around the agent.

For professional work, define:

• Work product: the only artifact the agent may change.

• Fixed controls: named versions of the specification and verification plan.

• Environment: approved sources, tools, runtime, and permissions.

• Selection rule: how a passing candidate must improve on the retained version.

• Limits: iteration, time, or cost budget.

• Stops: success, plateau, unresolved evidence, or a decision reserved for a person.

The baseline is the first retained version. The loop never has to accept a worse candidate merely to keep moving.

Separate Findings From Decisions

The verifier returns evidence-based findings: MET, NOT MET, or UNRESOLVED. The loop then makes an operational decision:

DISCARD means restore the best retained version, not erase the experiment. The loop should preserve a compact lesson about what changed, what evidence appeared, and what should be tried next. Failed attempts can improve the next decision even when they do not improve the work product.

NASA describes a baseline as an agreed set of requirements, designs, or documents whose changes are controlled [2]. NIST similarly emphasizes documented, repeatable evaluation and a traceable basis for management decisions [3]. Together, these ideas explain why a loop needs fixed controls and a visible record, not merely repeated generation.

The Prompt

Run a bounded improvement loop on [WORK PRODUCT].

Fixed controls:
- Specification: [VERSION]
- Verification plan: [VERSION]
- Approved sources, tools, runtime, and permissions: [ENVIRONMENT]
- Selection rule: [HOW A PASSING CANDIDATE IMPROVES ON THE RETAINED VERSION]
- Success, plateau, and stopping conditions: [CONDITIONS]
- Iteration, time, or cost limit: [BUDGET]

Preserve the baseline as the first retained version. Modify only the work product.
Use a separate verifier role or fresh context to apply the approved verification
plan to every candidate.

For each iteration:
1. Choose one change using the retained version and prior lessons.
2. Produce a candidate version.
3. Verify it as MET, NOT MET, or UNRESOLVED with supporting evidence.
4. KEEP it only when all required conditions are MET and it satisfies the
 selection rule. DISCARD it when a condition is NOT MET or it does not improve.
 HALT when a material condition is UNRESOLVED.
5. Record the change, evidence, decision, lesson, and next action. Carry the
 lesson into the next iteration.

Continue without requesting approval until the budget ends or an approved
success or plateau condition is met. Halt if continued work would require
changing a fixed control or making a decision reserved for a person.

Return the best retained version, final verification report, iteration record,
unresolved issues, and proposed control changes for a future run. Do not apply
those proposed changes during this run.

What Makes the Loop Autonomous

The agent does not need a person to approve every attempt. It can choose a bounded change, generate a candidate, invoke the verifier, decide KEEP or DISCARD under the approved rules, record the lesson, and continue for the allowed number of iterations.

Human authority moves to the boundaries. Before the run, a person approves the objective, controls, environment, budget, and escalation conditions. After the run, a person reviews the best retained version, unresolved issues, and any proposed changes. During the run, human intervention is required only when a material issue is UNRESOLVED or the next action would exceed those boundaries.

This is what turns a sequence of prompts into an actual loop: state persists across iterations, lessons influence the next attempt, and the best verified result is protected.

Questions

Can the loop improve the specification or verification plan?

It can propose changes, but it should not apply them during the active run. Approve new control versions first, then begin a new run so every candidate is judged on the same basis.

Does DISCARD mean the iteration was wasted?

No. The candidate is not retained, but its evidence and lesson remain in the iteration record. That learning should narrow the next change.

How many iterations should run without a person?

Use the smallest budget that can reveal useful improvement without creating unacceptable cost or risk. High-consequence work needs tighter limits and earlier escalation than reversible drafting work.

References

1. Karpathy, A. (2026). autoresearch. GitHub.

2. National Aeronautics and Space Administration. (2016). NASA systems engineering handbook: Appendix B—Glossary.

3. National Institute of Standards and Technology. (2023). Artificial Intelligence Risk Management Framework Core: Measure.

Notes

Join the EPM Network to access insights, influence our research, and connect with a community shaping the industry’s future.

Subscribe now

Support us by sharing this article with your friends and colleagues, or over social media.

A verification plan translates each material acceptance condition into an observable check. NASA defines product verification as proof of compliance with specifications, using test, analysis, demonstration, inspection, or a combination of methods [1]. The same principle works for a report, estimate, risk register, or decision brief: define the evidence before the result exists.

This post stands alone. You can use the method once for a single deliverable, even if no improvement loop follows.

Design the Check Before Production

For every acceptance condition, specify the evidence, permitted source, method, possible finding, and any judgment reserved for a person. The three findings are deliberately limited:

• MET: the available evidence supports the condition.

• NOT MET: the evidence shows the condition failed.

• UNRESOLVED: the evidence is insufficient or the decision belongs to a person.

UNRESOLVED is not a softer form of failure. It protects the review from false certainty when sources conflict, calculations cannot be reproduced, or professional judgment is required.

NIST recommends rigorous testing, uncertainty measures, benchmark comparisons, documented results, and independent review where appropriate. It also notes that independent review can reduce internal bias and conflicts of interest [2]. In practice, that means separating the producing role from the verifying role or, at minimum, starting the verification in a fresh context that does not inherit the producer’s self-assessment.

Use an Example for Form, Not Truth

An approved past example can show the expected structure, terminology, and level of detail. Language models can adapt from examples supplied in the prompt [3], which makes a good reference valuable.

But a reference is not evidence that the new work is correct. The verifier must check the new version against its approved specification and permitted sources. If the example conflicts with them, the conflict should be reported rather than copied.

The second AI is also not a final authority. Research on LLM judges documents position, verbosity, self-enhancement, and reasoning biases [4]. Independent AI review adds disciplined challenge; it does not replace accountable human approval.

Prompt 1: Design the Verification Plan

Design a concise verification plan for the work product defined in the attached
approved specification. Complete this before production begins.

For each material acceptance condition, define:
Condition | Observable evidence | Permitted source | Verification method |
Meaning of MET, NOT MET, and UNRESOLVED | Required human judgment

Use the attached approved example only to match the expected format, structure,
terminology, and level of detail. Do not treat it as evidence for this project.
Flag any conflict between the example, specification, or current sources.

Use UNRESOLVED when the evidence cannot support a finding or when a person must
make the decision. Ask me to correct or approve the plan. After approval, issue
it as a named version and stop before producing the work product.

Prompt 2: Apply the Approved Plan

Independently apply the attached approved verification plan to this completed
work product version. Do not revise it and do not rely on the producer’s
self-assessment.

For each acceptance condition, report:
Evidence inspected | MET, NOT MET, or UNRESOLVED | Finding | Required correction
or human decision

Cite the exact source, calculation, test, or passage supporting each finding.
Do not treat polish or similarity to the format example as evidence of
correctness. Report overall MET only when every required condition is MET, NOT
MET when any required condition is NOT MET, and UNRESOLVED when any material
condition remains unresolved. End with all required human decisions.

Keep Findings Separate From Actions

Verification reports what the evidence supports. It should not quietly rewrite the work product, relax the acceptance conditions, or decide whether a candidate should replace an earlier version.

For a one-time deliverable, a person can use the report to approve, revise, or reject the work. In a later bounded loop, the same findings can feed explicit KEEP, DISCARD, or HALT decisions. Keeping those decisions outside verification makes the review reusable and easier to audit.

For higher-consequence work, add requirement identifiers, source owners, calculation files, test environments, reviewer independence rules, and retained evidence. The basic method stays the same: condition, evidence, method, finding.

Questions

Must the verifier be a different AI model?

No. A different model can add diversity, but a separate role or fresh context is the minimum practical boundary. Critical findings still need the professional review required by the task.

Can the verifier fix problems as it finds them?

Not during the same review. Mixing correction and verification makes it difficult to know which version was actually checked. Report the finding first; revise and verify a new version separately.

What if a condition cannot be checked automatically?

Define the available evidence and return UNRESOLVED with the required human decision. A clear escalation is more useful than a fabricated pass.

References

1. National Aeronautics and Space Administration. (2016). NASA systems engineering handbook: Verification and validation definitions.

2. National Institute of Standards and Technology. (2023). Artificial Intelligence Risk Management Framework Core: Measure.

3. Brown, T. B., et al. (2020). Language models are few-shot learners. Advances in Neural Information Processing Systems, 33, 1877–1901.

4. Zheng, L., et al. (2023). Judging LLM-as-a-judge with MT-Bench and Chatbot Arena. Advances in Neural Information Processing Systems, 36.

Notes

Join the EPM Network to access insights, influence our research, and connect with a community shaping the industry’s future.

Subscribe now

Support us by sharing this article with your friends and colleagues, or over social media.

The first move is therefore not “write the report.” It is “help me define the work product.” A work product is the inspectable artifact created by the assignment: a briefing note, risk register, schedule narrative, design option comparison, or another concrete deliverable. It needs a form, an intended user, and a useful outcome.

A specification records the approved conditions for that work product. NASA’s systems-engineering guidance offers a useful distinction: a need should describe the problem rather than prematurely prescribe the solution, while a good requirement should be clear, feasible, unambiguous, and verifiable [1]. The same discipline applies to AI-assisted professional work.

Let the AI Interview You

Most people begin with an incomplete mixture of goals, constraints, preferences, assumptions, and solution ideas. A one-question-at-a-time interview slows the conversation just enough to separate them.

The interview should establish six things:

• Purpose: What outcome should this work product enable?

• Users: Who will read, use, approve, or be affected by it?

• Form: What exactly will be delivered?

• Scope: What is included, and equally important, what is excluded?

• Conditions: What must be true for the result to be acceptable?

• Authority: Which assumptions or decisions require a person’s approval?

This is not limited to software. Government digital-service guidance uses the same logic in a compact form: identify the actor, the need, the goal, and acceptance criteria that describe when the work is done [2]. In a construction or engineering context, the vocabulary may change, but the principle does not.

Keep Three Boundaries Clear

The specification defines the target. It does not produce the work product, design the verification method, or authorize autonomous iteration.

That separation matters. If the same conversation quietly moves from defining the need to choosing the solution, important assumptions can disappear into the draft. If it also decides how success will be checked, the criteria may be softened to fit what the AI has already produced.

The interview should end with a named, approved version. That version becomes a stable reference whether the next step is completed by a person, one AI, a team, or a later improvement loop.

The Prompt

Help me define a work product and its specification by interviewing me.

Ask one question at a time. Clarify the problem, who will use the result, the
useful outcome, the smallest worthwhile scope, and the constraints. Separate
needs from proposed solutions and confirmed facts from assumptions. Do not fill
material gaps yourself.

When you understand the task, summarize:
- Goal and intended use
- Work product, form, and intended users
- In scope and out of scope
- Requirements and constraints
- Acceptance conditions describing what must be true
- Assumptions, open decisions, and approval authority

Ask me to correct or approve the summary. After approval, issue the specification
as a named version. Stop there. Do not design the verification method or produce
the work product.

Start by asking me to describe the project idea in my own words.

What a Good Output Looks Like

A strong result is short enough to review and specific enough to govern later work. The work product is named in concrete terms. Scope boundaries are visible. Acceptance conditions describe outcomes rather than vague aspirations. Assumptions are not presented as facts. Open decisions have owners.

For low-risk work, a concise specification may be enough. For contractual, safety-critical, or multi-party work, add identifiers, source references, tolerances, traceability, and formal change authority. Those are extensions of the same model, not reasons to make every prompt heavy.

Once the specification is approved, the next question is no longer “what are we making?” It is “what evidence would show that this version meets the approved conditions?” That is the role of the verification layer.

References

1. National Aeronautics and Space Administration. (2016). NASA systems engineering handbook: Appendix C—How to write a good requirement.

2. Government Digital Service. (2016). Writing user stories.

Notes

Join the EPM Network to access insights, influence our research, and connect with a community shaping the industry’s future.

Subscribe now

Support us by sharing this article with your friends and colleagues, or over social media.

AI has become an important part of our research. We have explored practitioners’ expectations about its role in project work [1] and the implications of recent advances for engineering practice [2]. Within our group, we have also been collecting prompts and ideas to test in our own work.

We increasingly share these ideas with students and professional colleagues. PM Along AI gives them a public home where others can find them, try them, and adapt them to their own work.

A useful prompt does more than request an answer. It helps us define the task, examine the available inputs, make assumptions visible, and describe how a good result will be checked. Documenting why we use a prompt also helps others decide where it fits their work.

Prompting Idea

Map your work before deciding what to learn

Growing alongside AI means developing the expertise to direct its contribution and judge its results. That starts with understanding the work itself.

AI will not affect every task in the same way. A scheduler may use it to organize inputs or test construction logic. A risk analyst may use it to identify assumptions or compare scenarios. In both cases, the person still needs enough domain knowledge to notice missing evidence, weak reasoning, and project conditions that the AI does not understand.

Our work on project risk has highlighted the need to teach young engineers to challenge AI outputs [3]. Verification should be part of learning rather than a final administrative check. When we test an AI answer against evidence and professional knowledge, we also deepen our understanding of the task.

The following two prompts use that principle. The first creates a confirmed map of your work. The second uses that map to choose expertise and a practical experiment.

Prompt 1: understand my work

Interview me to understand how my work is done and where AI might help.

Ask one question at a time. Focus on the purpose of each task, how work
flows from inputs to outputs, where human judgment matters, how quality is
checked, and what repeatedly causes difficulty. Separate facts from my
assumptions.

Summarize what you learn as:
Task | Workflow | Human judgment and quality checks | Current difficulty

Ask me to correct the map before suggesting any use of AI.

Prompt 2: decide what expertise to build

Using my confirmed workflow map, identify where AI can assist now, what may
plausibly change over the next few years, and what human expertise should
grow alongside it.

Separate demonstrated capabilities from predictions and state the main
uncertainties. Recommend three learning priorities and one small experiment
I can try this week. For each recommendation, explain how I should verify
that it improves the work rather than merely making it faster.

Turn the result into practice

Choose one recurring task rather than trying to redesign an entire role. Correct the workflow map before asking for recommendations because a mistaken map will produce the wrong learning priorities.

Treat capability forecasts as scenarios, not facts. Tools, costs, organizational policies, and regulatory expectations can change. The useful question is which expertise remains valuable across several plausible futures.

Then run the proposed experiment on a low-consequence task. Compare the result with your current method. Check the quality, time required, errors, assumptions, and effort needed for review. Keep the practice only if it improves the work without weakening accountability.

This approach turns AI adoption into deliberate learning. The aim is not to predict every capability. It is to build enough understanding to make better choices as the technology and the work develop.

Notes

Join the EPM Network to access insights, influence our research, and connect with a community shaping the industry’s future.

Subscribe now

Support us by sharing this article with your friends and colleagues, or over social media.

Receive PM Along AI

To receive these posts, click Unsubscribe at the bottom of an email you have received from EPM Research. On the subscription preferences page, toggle PM Along AI on and save your preferences if prompted.

References

1. Zangeneh, P., & Ghorab, K. (2025, March 12). Transformative technologies in construction and project management: 2025 outlook and survey results. EPM Research.

2. Zangeneh, P. (2026, September 11). AI takes shape: What recent AI advances mean for architecture, engineering, and construction. EPM Research.

3. Zangeneh, P. (2025, November 27). Industry should brace for the first wave of AI-driven risk management failures. EPM Research.

Opportunities is a separate section of EPM Research dedicated to sharing job opportunities and hiring notices with our readers.

Keeping opportunities in their own section means readers who are not looking for a position are not flooded with notices, and readers who are looking do not have to search for them.

Subscribe and Turn Opportunities On

Subscribing to EPM Research does not automatically enroll you in Opportunities. The platform treats each section independently, and the Opportunities notification is off by default. To receive the posts in this section,

1. If you have not done that yet, first you need to subscribe to the EPM Research letters.

2. Open the welcome email you received, or any email you have received from EPM Research, and scroll to the very bottom.

3. Click the small “Unsubscribe” link under the copyright line. Despite the wording, this does not unsubscribe you, it opens your notification settings for this publication.

3. On the settings page, switch the “Opportunities” toggle on, so it shows green.

That is all, and the setting saves itself. Ignore the green “Unsubscribe from all” button; you do not need to press anything else.

Key Takeaways

• AI can now work with editable geometry through existing CAD and Modelling tools. Astra writes parametric programs, Fable executes long workflows, and Atlas captures physical environments in point clouds.

• A step change in a narrow domain needs two things: the data, and a harness that can act on it. AEC holds both.

• The durable capability is designing the operating loop around the model. A proper loop needs three layers: specs, verification, and the environment.

• One industry after another is now building verification layers from scratch. AEC spent thirty years building one called BIM.

• On paper, architecture and engineering is among the sectors most exposed to AI. In practice, few of us use it.

For thirty years the assumption was that AI would reach the construction industry through documents: specifications, RFIs, contracts, submittals. It did, with mixed success. But the announcements of early September 2026 point somewhere else, and they promise the moment the Architecture, Engineering and Construction (AEC) industry has been preparing for.

This one is about geometry. And it arrives at an industry that already owns what others are now scrambling to build. One industry after another is now building a verification layer from scratch. AEC spent thirty years building one and called it BIM.

On 1 September 2026, World Labs announced Atlas, an “omni world model” pretrained from scratch to operate natively on text, images, video and 3D, rebuilding a real environment from as few as two or three images, and exporting point clouds or 3D Gaussian splats [1]. Two days later, on 3 September, OpenAI shipped GPT-6 Astra, and for the first time a flagship general model’s launch material led with CAD and 3D work rather than text and code [2]. Alongside them, also on 1 September, Anthropic shipped Claude Fable 5.1, which made no 3D claim whatsoever. Its gains were in long-horizon agentic work [3].

In this post, I aim to review the following

1. Astra, Fable, and Atlas

   1. OpenAI GPT 6 Astra in 3D modeling

   2. Anthropic Fable 5.1 in BIM, scheduling, and shape files

   3. World Labs Atlas in point clouds

2. Why this might be AEC’s moment

   1. Change will come in major steps

   2. AI development triad: spec, verification, and the environment

   3. BIM as a verification layer

3. Sector specific AI developments

4. Human-AI collaboration mindset

5. Question then becomes

Let us dive in.

1. What Astra, Fable and Atlas actually do with geometry

Most coverage of the GPT-6 Astra launch described a model that builds 3D models, on the strength of a reported 95.9% on the BenchCAD leaderboard. The details behind that number matter.

BenchCAD is a test of 17,900 CAD programs across 106 industrial part families, roughly half anchored to published ISO, DIN, EN, ASME or IEC standards, every one compiled and run before it entered the set [4].

What it measures is narrow and real: does the model’s code execute, and does the shape it produces occupy the same space as the reference answer. What it does not measure is everything around the shape, including manufacturing tolerances, undocumented design intent, assembly information, and whether a part complies with the standard it was anchored to rather than merely resembling it [4].

A high score means a model can write a program scripts that runs in a modelling software to produce a geometric model on a test the vendor administered to itself. It does not mean the model understood why the part is shaped that way [5].

Independent CAD reviewers report models getting the broad shape right while failing parametric intent: missing detail, misreading engineering parameters, choosing the wrong modelling steps [6].

1.a. OpenAI GPT 6 Astra in 3D modeling

The demonstrations read more clearly than the scores. GPT-6 Astra drives Blender (3D modelling software) from a plain text prompt and prepares assets for Unreal Engine 5, and the game studio Playco built three themed playable prototypes from a single grey-box model, reporting 50% fewer manual fixes [7].

An interesting public example is the reconstruction of the Palace of Fine Arts, a full Blender scene built from reference imagery in one session [8].

The sequence below is roughly the process that the model followed.

1. Read the reference into a scene description. Astra takes the photographs as input and writes down what is in them: a domed rotunda, a colonnade, a peristyle wrapped around a lagoon, and the massing relationships between the three. Language, not pixels.

2. Fix the geometry as parameters. Rotunda radius, column diameter, bay spacing, cornice heights and lagoon extents become named values, so the scene can be rebuilt at a different scale without being redrawn.

3. Write the generator, not the model. The output is Blender Python. The walkway is a loop over a bay spacing rather than a row of individually placed columns, which is why the finished scene arrives as editable objects instead of one mesh.

4. Run it, read the error, edit the code. The script executes inside Blender and the model reads back what happened, whether a traceback or a scene that does not match what it asked for, and rewrites its own program. This is the long-horizon loop, and it is where most of the session goes.

5. Layer materials, texture and environment. Surfaces, the water plane and the surrounding planting are added as separate passes over the geometry the script has already placed.

6. Set a matching camera and render. A view is framed to match the reference photograph so the render and the source can be put side by side. It is the only check in the whole sequence performed by eye.

The third step is the one that matters. What Astra produced is a program that builds the Palace of Fine Arts, not a picture of it. Change the bay spacing and the walkway rebuilds. That is a different kind of artefact from a mesh.

1.b. Anthropic Fable 5.1 in BIM, scheduling, and shape files

Claude Fable 5.1 makes no CAD claim of its own. Anthropic’s launch material published no BenchCAD figure, but the 84.3% is its score on the BenchCAD leaderboard on a self-reported basis as Astra’s [5].

That the number is respectable anyway, and last only tow days at top of the leaderboard is worth a moment’s thought.

Complex modelling is rarely one act of geometric insight; it is a long chain of small ordinary decisions: pick the sketch plane, extrude, fillet, check the result, notice the wall went the wrong way, undo, try again. Parametric CAD punishes the first wrong step and rewards patience, so a model that can carry a set of constraints through a fifty-step edit session and recover from its own mistakes will out-model one that writes a more elegant program in a single pass.

Claude Fable 5.1’s primary reported gains sit in exactly that type of work: long-horizon agentic tasks where a model holds a goal across many steps, calls tools, reads what came back, and corrects itself rather than committing to its first answer [3].

Three demonstrations, I believe, could show what that looks like in our own work. Both were run on Claude Fable 5.1, each in a single session, with no human intervention between the opening instruction and the final file.

Photo → BIM

Asking Claude to create a BIM model from a picture. One street-level photograph in, an IFC4 model at LOD 200 out. And the model took the following seven steps:

1. Read the photo into primitives. Volumes, roof planes, openings, cladding zones, a scene description, not pixels.

2. Scale off standard components. E.g., Garage door 16 × 7 ft.

3. Fix a frame, parameterize. X/Y/Z convention; every dimension a named constant with its basis.

4. Build the IFC skeleton. Project → site → building → two storeys, metric, Body context.

5. Generate the elements. Swept profiles, not meshes, editable in native BIM tools.

6. Add openings, materials, provenance. Void/fill relationships, surface styles, confidence on every element.

7. Verify. Schema check, tessellate all 40 products, render against the photo. Caught the garage roof pitch.

BIM → Schedule

Asking Claude to create a construction schedule from the BIM model. The schedule was based on the geometry that was already in the model. Three steps:

8. Extract the inventory. GlobalIds, classes, volumes, the quantity basis and the join key.

9. Write the schedule as data. 70 activities, logic, crews, costs, each keyed to the elements it builds.

10. Run CPM, emit XER, verify. Working-day passes → 17 P6 tables → re-read with two parsers. Caught two dangling activities.

The last step of each run is the one worth noticing. The model checked its own output and found its own mistakes, a wrong garage roof pitch in the first, two dangling activities in the second, and neither was caught by a person. That is long-horizon behaviour doing work an engineer would otherwise do.

Calgary Rail Crossing Corridor

For the third example, I asked Claude Fable 5.1 to pull open geospatial data, built the shapefiles itself, and rendered the same rail network at different scales: the four subdivisions that meet at Alyth Yard, and the downtown corridor in detail.

That matters because scale is where most project geometry gets lost, since the network model, the corridor model and the structure model are usually three separate exercises. A system that can assemble the data and move between those scales inside one session is doing the work of coordination, not just the work of drafting.

1.c. World Labs Atlas in point clouds

World Labs’ Atlas does not appear on the same scale as is a different kind of system. It accepts text as a prompt but does not respond in text. The model reconstructs 3D scenes from as few as two or three images and exports point clouds or 3D Gaussian splats [1].

Atlas emits geometry directly, so a code-execution benchmark like BenchCAD has nothing to say about it. Reconstruction can be measured against ground truth, and World Labs reports a mean absolute-relative pointmap error of 25.3 against 28.7 for the next best system, a vendor-run result that has not been independently replicated [9].

Almost every project begins with a condition somebody has to measure, and today that means a scanner, a crew, a registration workflow in only few days. That is why existing-conditions models are skipped on small jobs and stale on large ones.

Point clouds and splats are not BIM, and nothing here authors a wall or a schedule. But as a source of geometric truth to check a model against, progressive capture is the input our clash, quantity and progress machinery has always lacked.

Read together, GPT-6 Astra, Claude Fable 5.1 and Atlas are not competitors on one axis. They are different answers to how a machine touches three-dimensional space: Astra writes the program, Fable runs the long errand, Atlas perceives the scene.

2. Why this might be AEC’s moment

2.a. Change will come in major steps

Capability in a narrow domain rarely improves along a smooth curve. It sits flat, and then it steps.

Language models playing chess is the clearest worked example. Performance across models and versions was erratic enough to look like a puzzle about reasoning, and the closest examination available traced most of it to two ordinary things: what was in the training data, and how the model was prompted and allowed to act [11]. Neither is a claim about intelligence and both are decisions someone made.

OpenAI has said nothing public about GPT-6 Astra’s training data composition, so what follows is inference rather than fact. But the same two levers are the plausible ones.

• The first is data: a deliberate injection of parametric CAD programs, Blender and Grasshopper scripts, IFC and USD structures, and 3D workflow traces into the training corpus.

• The second is the harness: computer use maturing enough to let latent spatial competence act. The model may always have half-known Blender; what changed is that it can now click.

The two are not alternatives. Data without a harness is knowledge the model cannot apply, and a harness without data is a model clicking through software it does not understand. A step change in a narrow domain generally needs both.

AEC owns an enormous body of structured, semantically rich, geometrically precise three-dimensional data: thirty years of coordinated federated models, as-builts, clash histories and fabrication models, much of it schema-conformant, and most of it authored by people who were paid and professionally liable to get it right.

2.b. The AI development triad: spec, verification, and the environment

One useful way to understand AI deployment is to divide it into three connected layers: write down what the tool must do, decide in advance how you will test that it does it, and build the surroundings it runs in. That is Spec, Verification, and Environment.

The division is not new, which is the strongest argument for it. Anyone who has specified engineering software has worked this way already.

The layers are connected rather than sequential: a weak spec makes a verifier impossible to write, and a poor environment makes a good spec unenforceable.

• Spec: a precise statement of what “done” means for a task, explicit enough that an agent does not have to infer the intent. Specs are built iteratively through use, not dumped upfront in a single document.

• Verification: an independent and ideally objective test of whether an output meets its spec. It answers accept or reject without a human forming an opinion. “Does it look right” is not a verifier.

• Environment: the persistent workshop an agent operates in, holding standing instructions, accumulated project knowledge, reusable skills and guardrails around destructive actions. It is what makes agent behaviour repeatable across tasks.

2.c. BIM as a verification layer

Our industry appear to have built the perfect verification layer in past 30 years.

A coordinated building information model is not a drawing, and it is not a rendering. It is a structured, queryable, rule-checkable assertion about a building, carrying clash detection, model-checking rule sets, schema validation against the Industry Foundation Classes standard [13], quantity extraction, and a research literature on automated rule-based code compliance that predates the current wave of models by well over a decade [14]. Each of those is a verifier in the technical sense.

This is why the text-only detail from Astras process matters. GPT-6 Astra’s contribution to AEC is not that it makes geometry. It is that it emits executable, checkable artefacts. Pair an agent that emits checkable artefacts with an industry that already owns the checker, and the loop closes.

If BIM is the verification layer, the next question is whether the sector is ready to use it as one. One thing for sure, the product activity is real: three AI-BIM products launched in June 2026 alone, Trunk Tools Cortex, Beam AI’s BIM CoPilot and OFA Group’s QikBIM, alongside Higharc’s US$95M Series C, which is a funding round rather than a product launch [15].

3. Sector specific AI developments

Two things happened this year that, read together, point at where AI in AEC goes next.

The first was a measurement. In March 2026 Anthropic published its read of the labour market [16]. Earlier studies asked what AI could do in a job. Anthropic asked what it is actually being used for. It looked at which work tasks people bring to Claude, then compared those with the tasks a model could in theory speed up. The first they call theoretical exposure. The second they call observed exposure.

Architecture and engineering has the widest gap between the two of any major occupational group. About 85% of our tasks are exposed in theory. Under 5% show up in the usage data. Business and finance sits at 94.3% in theory and 28.4% in practice

The second was a product. On 10 September 2026 OpenAI launched ChatGPT for Financial Services, built on GPT-6 Astra, with Morgan Stanley and Evercore as design partners. It connects to PitchBook, S&P Capital IQ, LSEG and Moody’s, and it produces research, models and pitchbooks in a firm’s own templates [23].

That is the vertical playbook: pick a sector with high exposure, connect the model to the data and the templates that sector already trusts, and build it with the firms that set its standard of work.

AEC fits that profile. We have the widest exposure gap of any major occupational group, a mature data layer in BIM and IFC, a flagship model whose launch material led with CAD [2], and a first wave of AI-BIM products aimed squarely at this work [15]. What we do not have yet is the usage. We track that side of it in our annual Transformative Technologies in Construction survey.

Sector-specific AEC systems are the predictable next moves.

4. Human-AI collaboration mindset

The collaboration between Human and AI is still in its infancy, and while it is, it pays to watch how others use it.

Reading what power users and whole organizations actually do with these tools is consistently more useful than reading what the vendors say they do, and the first thing you notice is how differently everyone works.

Anthropic’s own usage study makes the point at scale: what people ask for shifts by hour of the day, by day of the week and by occupation, and personal use climbs from about 35% of conversations on a weekday to roughly 50% at the weekend [19]. There is no single way to use one of these systems.

Commonalities do emerge, though, and they are the interesting part. The people getting the most out of these tools are not writing better prompts. They are building something that runs.

Which is how the vocabulary changed. Boris Cherny, who leads Claude Code at Anthropic, put it about as plainly as it can be put in June 2026: “I don’t prompt Claude anymore. I write loops and the loops do the work. My job is to write loops” [20].

The label matters less than what it concedes, which is that the leverage has moved from the model to the structure around it.

A second idea explains why the loop has to be built rather than assumed. Andrej Karpathy draws a distinction between animals and ghosts [21]. An animal arrives with drives it did not learn. It is hungry, it is curious, it wants things and pursues them without being asked. A large model has none of that. It is closer to a ghost summoned out of a corpus of human text, capable and inert at the same time, with no intrinsic interest in whether your drawing set comes out right. A system with no goals of its own will not notice that it has drifted, will not mind that the model no longer matches the specification, and will not stop. (This goes to the core of the alignment debate, which is a topic for another day).

A third idea is Ethan Mollick’s, and it is about timing rather than architecture: the AI you are using today is the worst AI you will ever use [22].

The model is the part that improves without your help, so building your work around this quarter’s model capability means building on the thing that will change, while building the loop around it means building on the thing that will last. A verifier written against IFC outputs this year runs unmodified against next year’s model, and returns better results, because the thing in the middle of the loop got stronger.

5. Question then becomes

With all the buzz about AI tools, perhaps there is solace in the long, struggling hours it takes to shape their results into a useful artefact, such as this letter. Those hours invite deeper thoughts and questions about how we should react to this new invention.

• One major thought I had is this: if verification makes autonomy possible, could it also become a ceiling? If we delegate only what can be measured, will we eventually value what is checkable more than what is important? And when that happens, what will be important?

• If AI performs the repetitive work through which young professionals once developed intuition, can the industry preserve expert judgment without preserving the experiences that created it?

• Our codes and verification rules preserve lessons from the past. When AI proposes something genuinely new, should the verifier protect society from novelty, or protect precedent from being challenged?

• On paper, architecture and engineering is among the sectors most exposed to AI. In practice, few of us use it. Why is that? Is this a warning sign or an Aha moment.

Please share your thoughts.

Notes

Join the EPM Network to access insights, influence our research, and connect with a community shaping the industry’s future.

Subscribe now

Support us by sharing this article with your friends and colleagues, or over social media.

If you wish to share your opinion, provide insights, correct any details in this article, or if you have any questions, please email editor@epmresearch.com.

Refer to this article using the following citation format:

Zangeneh, P. (2026), “AI Takes Shape: What Recent AI Advances Mean for Architecture, Engineering, and Construction,” EPM Research, 11 September 2026.

Partner with EPM Research

EPM Research at the University of Calgary aims to be an effective research arm for Canadian Project and Construction Management industry.

Generative AI in construction and project management is a core track of our research. Our work ranges from AI adoption strategy to independent reviews of AI systems to pilot development and implementation. We place a particular focus on supporting small and medium-sized Canadian organizations in the public and private sectors, a priority encouraged by the Engineering Project Management Endowment and Canadian research funding agencies.

Please use this form to tell us what you're working on and how we might help. We review submissions regularly and will contact you when there is a good fit with our research directions and capacity.

Partner with EPM Research at UCalgary

About the author

Pouya Zangeneh, Ph.D., P.Eng., is Assistant Professor in the Department of Civil Engineering and Endowed Chair in Engineering Project Management at the Schulich School of Engineering, University of Calgary. He has more than 15 years of experience in multi-billion-dollar capital projects and capital investment planning across heavy industrial, utilities, energy, mining, and infrastructure sectors.

• ORCID: https://orcid.org/0000-0001-5074-302X ·

• Profile: https://profiles.ucalgary.ca/pouya-zangeneh

Related reading from EPM Research

• Industry Should Brace for the First Wave of AI-Driven Risk Management Failures

References

1. World Labs. (2026, September 1). Atlas: A world model for spatial intelligence. World Labs.

2. OpenAI. (2026). GPT-6 Astra: A new generation of intelligence. OpenAI.

3. Anthropic. (2026, September 1). Introducing Claude Fable 5.1 and Claude Mythos 5.1. Anthropic. With model specifications from llm-stats (2026) and CellCog (2026), Fable 5.1 pricing, context window and benchmarks. https://www.anthropic.com/claude-fable-and-mythos-5-1

4. BenchCAD: A comprehensive, industry-standard benchmark for programmatic CAD. (2026). arXiv:2605.10865.

5. BenchCAD. (2026). Official leaderboard, benchcad.com/leaderboard, including self-reported, with-tools entries flagged as not re-graded; and llm-stats. (2026). BenchCAD leaderboard mirror, which reports a differing set of figures. https://llm-stats.com/benchmarks/benchcad

6. NoriCAD. (2026). GPT-6 Astra and CAD: Could AI change 2D and 3D design?. NoriCAD.

7. Crypto Briefing. (2026). OpenAI’s GPT-6 Astra enables 3D modeling from Blender to Unreal Engine 5; and OpenAI. (2026). Playco: game prototyping with Astra. OpenAI customer story (source of the customer-reported 50% fewer manual fixes). https://cryptobriefing.com/openai-astra-3d-modeling-unreal-engine/

8. Shameem, S. (2026, September). Astra 3D reconstruction of the Palace of Fine Arts in Blender. X.

9. AlphaSignal. (2026). World Labs’ Atlas beats specialized 3D models with one omni model: reported sparse-view reconstruction and camera-control preference results. https://alphasignal.ai/news/world-labs-atlas-beats-specialized-3d-models-with-one-omni-model

10. OpenAI. (2026). GPT-6 Astra model card. OpenAI API documentation.

11. Dynomight. (2024). OK, I can partly explain the LLM chess weirdness now. Dynomight.

12. Wing Venture Capital. (2025). RL environments for agentic AI: Who will win the training and verification layer by 2030. Wing VC.

13. ISO 16739-1:2024. Industry Foundation Classes (IFC) for data sharing in the construction and facility management industries, Part 1: Data schema. International Organization for Standardization.

14. Eastman, C., Lee, J., Jeong, Y., & Lee, J. (2009). Automatic rule-based checking of building designs. Automation in Construction, 18(8), 1011–1033. https://doi.org/10.1016/j.autcon.2009.07.002

15. Geo Week News. (2026). AI BIM automation sees concentrated June activity with four major launches. Geo Week News.

16. Anthropic. (2026, March 5). Labor market impacts of AI: A new measure and early evidence. Anthropic. (Source of the observed-exposure measure, the 0.6-point BLS growth relationship, and the 22–25 hiring figures.)

17. Dezeen. (2026, March 11). Architects and engineers among professions most automatable by AI according to Anthropic. (Reports architecture and engineering at roughly 70% theoretical exposure with observed exposure close to zero.)

18. Monograph. (2026). Why architecture and engineering have the most untapped AI potential of any industry. (Reports the same gap at roughly 85% theoretical against about 5% observed, and the American Institute of Architects 2025 figures: 6% of practising architects using AI regularly, 8% of firms with AI in a process. The two write-ups differ; both are secondary coverage of the same report.)

19. Anthropic. (2026, June 26). Anthropic Economic Index report: Cadences. Anthropic. (Source of the weekday versus weekend usage split, the 93% artefact figure, and the note that construction and extraction occupations are under-represented.)

20. The New Stack. (2026, June 10). The Anthropic leader who built Claude Code says he ditched prompting — now he just writes loops. (Source of the Boris Cherny quotation.)

21. Patel, D. (2025, October). Andrej Karpathy: “We’re summoning ghosts, not building animals.” Dwarkesh Podcast. Primary source for the animals-versus-ghosts distinction and the “ethereal spirit entities” phrasing; see also Karpathy, A. (2025), Animals vs ghosts, and TeamDay (2025), We’re building ghosts, not animals, for secondary write-ups.

22. Mollick, E. One Useful Thing. Source of the “the worst AI you will ever use” framing, repeated across his writing

23. Open AI, (2026, September 10). Introducing ChatGPT for Financial Services. OpenAI.

We are looking for someone with practical experience in software architecture, Azure/cloud systems, Agile/Scrum workflows, DevOps practices, and the emerging paradigm of AI-assisted software development, including the use of generative AI coding tools and management of AI-generated code.

Expected involvement

• Few hours per week (flexible)

• Coaching students on Agile/Scrum practices

• Supporting software architecture and system design

• Guiding Azure DevOps / CI-CD pipeline development

• Assisting with Git workflows, testing, deployment, and documentation practices

• Helping establish code review and quality assurance practices for AI-generated code

• Advising on responsible use of AI coding tools (e.g., Copilot, Cursor, Claude Code) within a research and team context

• Helping establish continuous improvement processes across projects

Preferred experience

• Demonstrated, hands-on experience with generative AI coding tools and workflows (e.g., GitHub Copilot, Cursor, Claude Code, or similar)

• Cloud-based application development (Azure etc.)

• Azure DevOps or GitHub Actions

• Agile/Scrum project management

• Software architecture and API/system design

• DevOps and CI/CD pipelines

• Experience managing, reviewing, and governing AI-generated code in a team or production setting

• Experience mentoring junior developers or student teams is an asset

This role is ideal for a senior developer, DevOps engineer, or software architect with demonstrable, hands-on experience in AI-native development and a genuine interest in applied AI, engineering systems, and mentoring young researchers in a research-driven product environment.

Apply with email

To express interest, please send a short introduction, a summary of relevant experience, and a two-page CV to this email address with the subject line: "AI Native DevOps Coach - Your Name"

Takeaways

• The construction workforce gap is not just a shortage of workers; it is a mismatch between volatile demand, delayed supply, and short-term fixes that erode productivity.

• Industrial booms and maintenance shutdowns create predictable labour spikes, but at different rhythms, and the industry lacks institutional mechanisms to coordinate them.

• Productivity is the hidden multiplier: when productivity falls, the same project pipeline requires more labour hours, more supervision, and more rework.

• Apprenticeship intake alone will not solve the gap if completion remains low or workers exit during downturns.

• The next demand wave is arriving, this time with stronger centralized national drivers.

I was invited to speak at the Construction Innovation Centre Forum, in a session on Canada’s construction workforce gap, and to participate in a panel discussion on demand, supply, and productivity, with representatives of the Government Alberta, Alberta Building Trades, and Alberta Centre for Labour Market Research (ACLMR).

That invitation became an opportunity for me to pull together what I have learned over the past few years about construction workforce issues by speaking with industry veterans in Alberta and across Canada, reviewing the literature, and trying to understand the system-level forces shaping the problem.

This article is my attempt to put those reflections in writing. Much like the panel discussion itself, it is in the spirit of contributing to a difficult and important conversation.

Below are the main sections of the article.

1. Defining the Gap

2. The Industrial Boom-Bust Cycle

3. The Heavy Industries Maintenance Cycle

4. The Coordination Gap: Everyone Plans Alone

5. Supply Cannot Respond Fast Enough

6. The Band-Aid Trap: Why Usual Fixes Can Backfire

7. Productivity: The Hidden Multiplier

8. Unions: Stabilizers in the Boom, Exposed in the Bust

9. The Understanding Gap

10. The Next Wave and Our Chance to Course-Correct

11. Question then Becomes

The Canadian construction sector is fortunate to have many people across industry, government, unions, training institutions, and academia who are working, sincerely and every day, to strengthen its health and maturity. My hope is that our research at the University of Calgary can also help with tangible impacts on the ground. The stakes are high, and so is the potential for positive impact.

To this end, our EPM Research Group at the University of Calgary is beginning a project titled “Synthesis and Management of Regional Construction Project Portfolios.” Our angle on the problem is to understand how overlapping projects unfold, how regional labour pools and productivity constraints interact, and how better coordination might help. Please feel free to reach out if this work is of interest or if you would like to participate as an expert in this program.

Let us dive in to review the problem.

1. Defining the Gap

The construction workforce crisis is often described as a shortage: too much work, not enough workers.

That framing is not wrong. Owners, contractors, and project managers feel the shortage directly when bids come in higher, crews are hard to staff, schedules slip, or experienced supervisors are stretched across too many sites. But the simple shortage story misses the deeper mechanisms.

At its core, the workforce gap can be expressed as: Workforce Gap = Effective Labour Demand − Available Labour Supply. In a more explicit form: G(t) = D(t) / P(t) − S(t)

Where if D(t) is the volume of construction demand, it will be adjusted by P(t), that is the productivity, to arrive at the effective labour demand.

When productivity falls, the same amount of construction output requires more labour hours, more supervision, and often more rework. In practical terms, lower productivity increases effective labour demand even if the project pipeline stays the same.

This main workforce equation can be depicted as the system below, using a simplified system dynamics convention: green arrows indicate an increase, and red arrows indicate a decrease. We will use this convention throughout the post.

BuildForce Canada projects a shortfall of 108,300 workers by 2034, accounting for retirements and expected new entrants [1]. RBC has estimated that Canada may need more than 500,000 additional construction workers by 2030, largely because of housing and infrastructure demand [2]. The difference between these figures reflects differences in scope, assumptions, time horizon, and whether the estimate captures broader demand pressures and productivity effects.

While the exact numbers are debatable, the pressure is already visible across the industry. One survey reported that 72% of British Columbia contractors had turned down work because they could not secure enough skilled trades [3].

This framing also treats workers as interchangeable units of supply. A qualified tradesperson cannot be moved instantly from one project to another. Assuming labour supply as a pool of interchangeable workers is a useful modelling simplification. But it flattens the differences between trades, levels of experience, regional relationships, and career trajectories.

2. The Industrial Boom-Bust Cycle

BuildForce Canada estimates that of roughly 1.3 million construction workers in Canada in 2024, about 680,000,or 52%, worked in non-residential construction [1].

Industrial project activity is geographically concentrated in resource rich regions. Natural Resources Canada’s inventory shows that Canada had 504 major natural resource projects planned or under construction in 2024, worth a combined $632.6 billion [7]. These projects cluster in resource-rich regions and often draw from overlapping pools of trades, supervisors, camp staff, and specialized contractors.

When commodity prices rise, many owners read the same signal at the same time. Projects that were shelved or marginal suddenly become viable. Firms restart work, contractors mobilize, and labour demand spikes.

Labour is one of the largest and most inflation-sensitive components of construction cost, though its share varies by project type, delivery model, region, and sector. In many projects, labour can represent roughly 20–50% of total cost. A 10–20% increase in labour rates can therefore materially affect project economics, especially when escalation also affects materials, equipment, accommodation, and subcontractor pricing [4].

The causal chain is straightforward: higher commodity prices make more projects economic; synchronized project activation increases demand for trades, supervisors, equipment, camps, and contractors; labour shortages push wages upward; higher wages increase total project costs; and marginal projects become uneconomic again.

Few robust projects survive the boom with good enough economics that buries bad decisions. Others are delayed, scaled back, or cancelled. Demand eventually falls. The cycle takes, typically, 3 to 7 years.

The human cost is equally important. Research on unconventional oil and gas boom-bust cycles shows a recurring pattern: employment and earnings rise during the boom, but unemployment, poverty, income losses, and welfare receipt can rise during the bust [5]. Firms often hire aggressively during booms, compete for the same experienced workers, and then release workers when the cycle turns [5].

Workers remember this instability. A young person considering the trades does not see only the wage premium at the peak. They may also see layoffs, travel disruption, camp work, family strain, and physical demands tied to volatile investment cycles.

The Alberta oil sands cycle illustrates the mechanism. Investment accelerated through successive peaks in 2008, 2012, and 2014. At the height of the boom, camp populations reached 47,000 workers, and the regional population had grown 141% in nine years. The 2015 oil price collapse then reversed the cycle: investment fell 45%, projects were delayed or cancelled, and the boom era had largely ended by 2018 [6].

Counter-cyclical examples do exist. IAMGOLD’s Côté Gold project in northern Ontario offers one case of more counter-cyclical project preparation. But such examples remain exceptions. The dominant pattern is still that owners pause together, restart together, and compete for labour together.

This is the first major feedback loop that operates on 3 to 7 years cycles: commodity signals synchronize demand, synchronized demand increases labour pressure, labour pressure increases project cost, and rising project cost eventually destroys part of the demand that created the shortage.

3. The Heavy Industries Maintenance Cycle

Heavy industries, such as, oil refineries, petrochemical plants, potash mines, pulp mills, and processing facilities, require periodic Shutdowns, Turnarounds, and Outages (STOs) for inspection, maintenance, and regulatory compliance. These events are usually planned months, and often years, in advance. Yet they can still create acute regional labour shortages.

A single outage may require hundreds of additional workers for only one or two weeks, compressing months of maintenance into a narrow execution window [8]. During that period, workers may face long shifts, night work, consecutive days without rest, and unfamiliar site conditions [8]. Since most operators do not carry enough permanent staff to execute shutdown work, they rely on multiple contractors at once, drawing heavily from the same regional craft labour pool [8].

Spring and fall are preferred turnaround windows, especially in northern climates where weather affects execution risk. When several facilities schedule STOs in the same season, they compete for the same trades at the same time.

Alberta’s 2025 turnaround season illustrates the pressure: refinery and plant maintenance overlapped with major capital projects, including Dow’s $8.9 billion chemical complex and Air Products’ $1.6 billion hydrogen facility, creating an acute crunch for pipefitters, crane operators, and scaffolders amid 3.6% sectoral wage inflation [9].

Shutdown work is time-bound and operationally critical (inelastic demand). A single day of schedule delay can mean the loss of a full day’s profit for these plants, which can reach millions of dollars. Therefore, when labour is scarce, firms often respond with high premiums, accelerated hiring, and in some cases, Harley-Davidson raffles.

This maintenance cycle also is connected to the industrial boom-bust loop. New resource projects increase future demand for processing, refining, and maintenance. When commodity prices are high, operators may run plants harder, which can increase maintenance intensity.

The geography reinforces the problem. Refineries, petrochemical plants, mineral-processing facilities, and upstream resource projects often sit in the same corridors: northern Alberta, northern British Columbia, Saskatchewan, northern Ontario, and coastal LNG and petrochemical clusters. They draw from overlapping pools of trades, supervisors, scaffolders, crane operators, and specialized contractors.

4. The Coordination Gap: Everyone Plans Alone

In many industries, competitors share some forms of infrastructure while still competing in the market. Airlines rely on air-traffic control. Retailers depend on shared logistics systems.

Construction and heavy industry have not developed an equivalent mechanism for regional labour-demand visibility partly because construction projects are highly unique, location-specific, and variable in their timing, scale, and labour requirements.

Firms do not usually share project pipeline, turnaround, or labour-demand data. In some cases, direct coordination among competitors over scheduling, capacity, or output may raise competition-law concerns in Canada and the United States [11].

Two refineries in the same region cannot legally agree to stagger their turnaround dates with each other, even though doing so would smooth demand for scaffolders, pipefitters, and electricians across the year.

That caution is understandable. But the absence of any lawful coordination mechanism leaves the system somewhat blind, and prone to opportunistic behaviours.

The result is predictable. Each firm plans from its own signals: its own project approvals, maintenance windows, procurement dates, and hiring needs. When commodity prices rise or regulatory shutdowns approach, many firms in the same region respond at once. Individually, each decision may be rational. Collectively, they create synchronized labour spikes.

5. Supply Cannot Respond Fast Enough

At the heart of this problem is a question many young people ask themselves: is construction a rational career choice?

The decision to enter and remain in construction depends on more than wages and short-term employment availability. It depends on whether the industry can offer a safe working environment, reliable income, reasonable benefits, career progression, and working conditions that do not systematically shorten a life.

In theory, the system has a stabilizing mechanism. When demand rises and wages increase, construction should become more attractive. More workers should enter apprenticeships, and supply should eventually recover. But in practice, that balancing loop is weakened by three problems: training delay, pipeline leakage, and career volatility.

First, supply arrives late. It can take three to five years for an apprentice to become a qualified journeyperson. By the time higher wages translate into skilled labour on site, the project cycle may already have changed.

Second, the pipeline leaks. More than 100,000 new apprentices registered in Canada in 2024, but only about 34,000 completed, a national completion rate of 19.9%, below pre-pandemic levels [12]. The barriers are well known: employers may hesitate to sponsor apprentices through training blocks, while apprentices already earning close to journeyperson wages may be reluctant to take the pay cut required for classroom training [12]. Ontario alone faces a projected shortfall of 52,000 construction workers by 2034, even with planned training investments [13]. Some skilled workers also leave for the United States, where wages, regulation, and immigration pathways may appear more attractive [12].

The federal government’s 2026 commitment of $6 billion over five years to train 100,000 new workers is therefore important, but not sufficient. It increases intake, but it does not by itself fix completion, retention, or mobility leakage [14].

Third, volatility makes construction less attractive as a long-term livelihood. Boom-time wages can be powerful, but what often follows is higher unemployment, poverty, income losses, and reliance on social assistance during the bust [5].

Firms respond rationally but narrowly. US construction wages rose by an average of 20% from 2021 to 2023 during the infrastructure boom, with some specialized trades seeing annual increases of 9% to 11% in high-demand regions [15]. During booms, they hire aggressively, compete for the same experienced workers, and push wages and costs higher [5]. When retention becomes the crisis, they add base pay, benefits, and overtime policies, reported by 61%, 63%, and 54% of firms respectively, but these responses raise cost without addressing the volatility that drives people away [16].

The result is a high-turnover workforce. Construction turnover exceeds 21% annually, with average tenure of about four years [16]. Replacing one worker can cost 50% to 200% of that worker’s annual salary [17]. The consequences are familiar to anyone managing projects: delayed schedules, compromised safety, higher costs, and diminished quality [17].

This is how volatility becomes generational. Workers who experience repeated busts may not return. Apprentices laid off during downturns may not re-register. Workers who leave remote project regions for cities may not come back. Parents, guidance counsellors, and experienced tradespeople then pass on the lesson: construction can pay well, but it may not be stable.

That perception matters. For every five Baby Boomers retiring from skilled construction roles, only two younger workers are entering to replace them [18]. One survey found that only 3% of young adults named construction trades as a career they wanted to pursue [19]. Only 38% of Gen Z believe skilled trades offer the best job opportunities, compared with 59% of Baby Boomers [18]. Workers under 25 make up 13.6% of the total US labour force but only 10% of the construction workforce [20]. Among young tradespeople, 42% cite social stigma and 35% cite lack of parental support as barriers to entering the trades [21].

6. The Band-Aid Trap: Usual Fixes Can Backfire

The usual responses to the construction workforce shortage, immigration pathways, out-of-province recruitment, apprenticeship subsidies, wage increases, and diversity campaigns are necessary. Large projects depend on mobile labour, and Canada will need both domestic and international workers to meet future demand.

But these fixes are structurally incomplete. They add inflow without necessarily reducing volatility, improving retention, fixing completion, or protecting productivity.

The most common short-term response is to widen the labour pool quickly: bring in workers from other provinces, recruit from abroad, pay premiums, and hire less experienced labour when experienced trades are unavailable.

In a tight labour market, the experienced workers who should be mentoring apprentices or integrating new crews are already stretched. The result is a familiar project-management trap: headcount rises, but productivity declines.

Canada’s 2026 fast-tracked permanent-residence pathway for up to 33,000 temporary workers in in-demand sectors, including construction trades, may help increase available labour [22]. But immigration alone does not solve the regional timing problem that causes shortages to appear suddenly in specific labour markets. Nor does it address the volatility that drives workers out of the industry, and the remote regions, after the boom.

Wage increases have the same limitation. Higher wages can attract workers during a boom, but they also increase project costs. When cost escalation pushes marginal projects into delay or cancellation, the wage response reinforces the boom-bust cycle it was meant to solve.

Training investments perhaps is the most positive strategy for the system, but without a holistic plans face a different bottleneck. More apprenticeship registrations do not automatically produce more journeypersons. With Canada’s apprenticeship completion rate at 19.9%, much of the investment in recruitment, employer sponsorship, and classroom training fails to become certified capacity [12].

The deeper issue is timing. The major feedback loops operate on different clocks: boom-bust cycles over three to seven years, apprenticeship formation over three to five years, retention and attrition over several years, and generational perception over one or two decades.

7. Productivity: The Hidden Multiplier

The usual workforce fixes share a hidden risk: they may add headcount without increasing effective capacity. A larger but less experienced workforce is not the same as a more productive workforce.

The workforce problem is often framed as a shortage of people. But productivity changes the size of the shortage itself. Formally, effective labour demand equals construction demand divided by productivity: D(t) / P(t). If productivity falls, the same volume of construction output requires more labour hours. If productivity improves, the same workforce can deliver more.

This distinction matters. If productivity falls by 30%, labour required for the same output rises by more than 30%, roughly 43%, because the system is now operating at 70% of its previous output per worker.

The evidence is troubling. Between 1970 and 2024, aggregate US labour productivity more than doubled, while construction labour productivity declined by roughly 30% [23]. Globally, construction productivity improved by only 10% from 2000 to 2022, compared with 50% across the total economy and 90% in manufacturing [24]. If this trajectory continues, the world faces a projected $40 trillion shortfall in meeting construction demand by 2040 [24].

The workforce shortage also reinforces the productivity problem. Tight labour markets push employers to hire less experienced workers, and construction productivity research identifies lack of worker experience as one of the strongest drivers of productivity loss [25].

Less experienced crews often require more supervision and generate more rework. Rework alone can reduce labour productivity by up to 30% and consume 5% to 9% of total project costs [26]. Sixty percent of general contractors and 68% of trades workers identify poor coordination and schedule management as key drivers of lost productivity [27].

This creates a reinforcing loop: labour shortage leads to less experienced crews; less experienced crews increase rework, supervision burden, and delays; lower productivity increases effective labour demand; and higher demand deepens the shortage.

8. Unions: Stabilizers in the Boom, Exposed in the Bust

Unions connect recruitment, training, wages, safety, mobility, benefits across multiple employers and projects.

When demand surges, owners and contractors look to unionized labour markets for trained workers, known wage structures, apprenticeship pathways, safety expectations, portable benefits, and institutional memory. Statistics Canada’s dedicated Construction Union Wage Rate Index is itself a reminder that negotiated wage structures remain a significant part of how construction labour markets are organized [28].

This stabilizing role matters. Collective bargaining coverage is associated with more formalized employment conditions, including wages, benefits, and job security [29]. Canada’s Building Trades Unions represent more than 600,000 skilled trades workers and report more than $300 million in annual private-sector investment across 197 union training centres [30].

But this creates a serious contradiction. During the boom, the industry asks unions to help absorb demand, mobilize skilled labour, and prevent wage competition from becoming pure spot-market chaos. During the bust, that same workforce is often left exposed. Owners defer capital. Contractors demobilize. Union workers return to out-of-work lists. Apprentices lose hours. Some firms shift back toward non-unionized labour at much lower rates. Training centres continue preparing people for a demand cycle that may already have collapsed.

Canada is already struggling to convert apprenticeship intake into certified capacity. In 2024, more than 100,000 new apprentices registered, but only about 34,000 completed; if nothing changes, Canada faces a persistent gap of more than 20,000 skilled trades workers per year [31].

If the industry calls on unions when labour markets overheat but abandons workers when demand falls, it should not be surprised when apprentices leave, journeypersons do not return, and the next generation questions whether construction is a career worth entering.

8. The Understanding Gap

There is another gap beneath the workforce crisis: a gap in shared understanding.

Economists, project managers, industry associations, governments, tradespeople, and unions are often looking at the same problem through different lenses and models. Each lens is useful but they each also has a blind spot.

Policy makers often view the problem through an economic lens: capital investment creates labour demand, demographics shape labour supply, and the difference becomes the gap. This framing is valuable, but it can smooth over boom-bust cycles, and treat productivity as relatively fixed.

Project managers experience the same gap as schedule and staffing risk. The problem is not a national shortfall in 2034; it is a pipefitter needed in Q3, a subcontractor that cannot mobilize, or a crew that arrives without enough experienced supervision. This frame is practical and urgent, but it often responds to symptoms rather than the system producing them.

Tradespeople and unions see the gap through a third lens: livelihood stability, safety, dignity, craft identity, and whether a life in construction is sustainable. From this perspective, productivity is not an abstract output-per-worker measure. It is shaped by site conditions, tools, supervision, fatigue, rework, and whether quality work can be done at a reasonable pace.

The policy consequences are significant. A policy solution may emphasize training and demographic projections. A project-management solution may emphasize schedule buffers, redundancy, and wage premiums. A labour solution may emphasize safety, working conditions, benefits, and retention. None of these is wrong. But each is incomplete on its own.

There is also a missed angle: the complexity and interconnectedness of how engineering projects actually unfold.

As with any complex, multidisciplinary problem shaped by noisy feedback, uncertainty, and delayed consequences, different actors see different parts of the workforce system. Economists see labour-market imbalance. Project managers see delivery risk. Tradespeople see the lived reality of work. Much like the old fable of the blind men and the elephant.

9. The Next Wave and Our Chance to Course-Correct

Another demand wave is forming, this time with a more centralized national driver.

The Government of Canada’s Major Projects Office has referred 17 nation-building projects for federal assessment and consultation, spanning energy, minerals and metals, and forest-sector projects [35].

A second pressure is emerging from AI and data-centre infrastructure. Data centres alone are projected to consume up to 9% of total US electricity generation by 2030, creating a new source of sustained construction and power-infrastructure demand [32].

There are also some encouraging signs with regards to new entrants. The Wall Street Journal’s description of Gen Z as the “Toolbelt Generation” pointed to rising trade-program enrolment, including a 23% increase in trade programmes and a 16% rise at vocational community colleges [18].

This is also where diversity and inclusion can become part of the workforce equation, not a separate social objective. The industry has a labour shortage while large parts of the adult population remain underrepresented in skilled construction roles. In the United States, women represent 11.2% of construction workers, but only 3.3% of workers in skilled construction trades and 8.5% of construction management positions [33]. Black and Asian workers are also underrepresented in construction, while Hispanic workers are often concentrated in labourer and helper roles rather than certified skilled-trade positions [33].

There are precedents for governmental coordination of workforce when it comes to infrastructure projects.

Earlier federal infrastructure programmes did not solve construction cyclicality, but they did create a more visible and predictable public-spending pipeline. The Investing in Canada Plan, for example, was explicitly designed as long-term, sustained funding to support planning and prioritization across orders of government [34].

Organizations such as BuildForce Canada then helped translate construction investment, labour-market conditions, retirements, mobility, and trade-level demand into a shared forecasting system used by owners, contractors, labour groups, governments, and training providers [1].

The opportunity now is to widen that logic beyond public infrastructure alone. If the Major Projects Office is already positioned as a point of contact for nationally significant projects, with a mandate to coordinate federal approvals, reduce duplication, structure financing, and reduce project risk, it could also help Canada see the cumulative labour-demand profile of the full project portfolio before the next shortage peaks [35].

10. Question then Becomes

Canada’s construction workforce gap is not simply a shortage of workers. It is produced by feedback loops: volatile demand, delayed training, pipeline leakage, productivity loss, and weak coordination. If that diagnosis is right, the response cannot be another isolated programme.

The question, then, is what an effective systemic intervention would look like, and how different institutions are positioned to design it.

Should it begin with visibility? A regional labour-demand observatory could help owners, contractors, unions, training providers, and governments see anonymized aggregate demand, by trade, region, quarter, and project phase, before the shortage arrives.

When no institutional coordination mechanism exists, what behaviours fill the vacuum, labour hoarding, last-minute poaching, inflated premiums, strategic overbooking, or quiet side arrangements that may not withstand scrutiny?

At what scale should this happen? The workforce gap persists partly because every firm optimizes locally inside a system no one can see.

And if the generational loop is beginning to turn, if rising Gen Z trade-program enrolment is real, what must the industry do now to meet that generation on its way in?

Readers familiar with my work on project management strategy may ask: what if this is a prisoner’s dilemma, where every firm has an incentive to defect anyway? But the deeper issue is how to change the game itself, from competition over a shared labour pool to a coordination game in which cooperation becomes rational.

Notes

Join the EPM Network to access insights, influence our research, and connect with a community shaping the industry’s future.

Subscribe now

Support us by sharing this article with your friends and colleagues, or over social media.

If you wish to share your opinion, provide insights, correct any details in this article, or if you have any questions, please email editor@epmresearch.com.

Refer to this article using the following citation format:

• Zangeneh, P. (2026). “The Workforce Gap: Demand, Supply, and Productivity.” EPM Research Letters.

References

[1] BuildForce Canada (2024). Construction and Maintenance Looking Forward: 10-Year Outlook for the Canadian Construction and Maintenance Industry. BuildForce Canada.

[2] Royal Bank of Canada (RBC) (2023). Carving a new path: How to solve Canada’s housing supply challenge

[3] Berto Contractors Ltd. (2025). Tackling Labour Shortages in BC’s Construction Industry

[4] Cleopatra Enterprise (2024). The Impact of Inflation on Construction Project Budgets

[5] Raimi, K., et al. (2022). A review of community impacts of boom–bust cycles in unconventional oil and gas development. Energy Research & Social Science.

[6] AInvest (2025). Rising Output and Labor Constraints in Canadian Oil Sands: A Strategic Opportunity Amid Structural Challenges

[7] Natural Resources Canada (2024). Natural Resources: Major Projects Planned or Under Construction, 2024 to 2034.

[8] Prometheus Group / Cleopatra Enterprise (2024). What are Shutdowns, Turnarounds, and Outages (STOs)?

[9] OilPrice.com (2024). Skilled Worker Shortage Stalls U.S. Construction Boom in 2024

[10] Autodesk / FMI (2024). 100+ Construction Industry Statistics. Autodesk Construction Cloud.

[11] Competition Bureau Canada (2023). Competitor Collaboration Guidelines; U.S. Federal Trade Commission & U.S. Department of Justice (2000). Antitrust Guidelines for Collaborations Among Competitors.

[12] CBC News (2025). In the trades, apprentices are increasing, but not enough are making it to certification. CBC News.

[13] EDI Weekly / Ontario Construction Secretariat (2024). Shortage of skilled labour facing Ontario construction industry: report. Engineered Design Insider.

[14] HPAC Magazine (2026). Feds announce $6B five-year Team Canada Strong plan to boost skilled trade apprenticeships

[15] CBRE / NAIOP (2022–2023). Construction Cost Challenges Shift from Materials to Labor. NAIOP — Commercial Real Estate Development Association.

[16] Bridgit (2026). 50+ Construction Workforce Retention and Turnover Statistics for 2026

[17] TradeWorx USA (2025). Building Tomorrow’s Workforce: Retaining Skilled Trades in a Competitive Market

[18] For Construction Pros / Wall Street Journal (2024). Gen Z Revitalizes the Skilled Trades Amid Construction Labor Shortages

[19] National Association of Home Builders (NAHB) (2017). NAHB Young Adult Survey on Career Preferences (cited in JCHS Harvard). Joint Center for Housing Studies, Harvard University.

[20] X3 Tradesmen (2024). The Shortage of Millennials & Young People in Construction Trades

[21] DEWALT / WorldSkills International (2024). Nearly Half of Young Tradespeople Say Social Media Inspired Them to Choose a Career in the Trades. PR Newswire.

[22] Singleton Law Firm (2025). Canada’s New Immigration Policies Targeted at the Construction Sector

[23] Gottlieb, R. L., et al. (2022). The Strange and Awful Path of Productivity in the U.S. Construction Sector. NBER Working Paper 30845.

[24] McKinsey Global Institute (2024). Delivering on construction productivity is no longer optional. McKinsey & Company.

[25] Al-Momani, A. H., et al. (2025). Critical Factors Affecting Construction Labor Productivity: A Systematic Review and Meta-Analysis. Buildings (MDPI).

[26] Gonzalez, V., et al. (2020). Study of relationship between rework and labor productivity in Building Construction Projects. Revista de la Construccion.

[27] ConstructConnect (2025). Construction Labor Market Remains Stagnant Amid Slowing Demand

[28] Statistics Canada. (2026). Construction union wage rate index, monthly. Statistics Canada.

[29] Statistics Canada. (2024). Collective bargaining coverage rate, 2023. Quality of Employment in Canada.

[30] Canada’s Building Trades Unions. (n.d.). Canada’s Building Trades Unions.

[31] Government of Canada. (2026). Chapter 2: Benefitting Canadians: A Canada for all. Spring Economic Update 2026.

[32] US Bureau of Labour Statistics (2025). What’s behind the projected construction employment growth from 2023 to 2033?

[33] BigRentz / ECI Solutions (2024). Women in Construction: The State of the Industry in 2024

[34] Housing, Infrastructure and Communities Canada. (n.d.). Investing in Canada Plan – Building a Better Canada. Government of Canada.

[35] Major Projects Office. (2026). Project Support. Government of Canada.

Last month, we had the pleasure of hosting a special session at the University of Calgary featuring Dr. Terence P. McNulty on 25 Years of Commissioning and Ramp-Up Performance: McNulty Curves Revisited.

The session brought together about 100 attendees, including a strong turnout from Calgary’s downtown professional community, and it sparked a thoughtful discussion on one of the most important yet often underestimated phases of major projects: the transition from commissioning to stable operations.

The session built on Dr. McNulty’s long-standing work on ramp-up performance and the updated McNulty Curves, which continue to provide a powerful lens for understanding why some projects achieve design capacity quickly while others struggle for months or even years.

Several messages from the session stood out clearly. First, strong ramp-up performance is rarely accidental. Second, the quality of commissioning and start-up planning matters enormously. Third, technical readiness alone is not enough. As the lecture emphasized, practical judgment, or phronesis, remains essential in navigating real project conditions, leading to my personal favorite slide:

For those who would like to revisit the material, the presentation slides will be available for a limited time here:

Session Slides – Limited-Time Link

Despite some last-minute difficulties caused by severe weather, which led to Terry’s flight being cancelled, we were still able to successfully hold the session online and ensure the event went ahead. We continued the evening with the reception dinner afterward, which gave attendees a valuable opportunity to connect, reflect on the discussion, and enjoy the evening together.

We would also like to sincerely thank the organizations that supported this event: the Canadian Institute of Mining, Metallurgy & Petroleum (CIM), Hatch, and the AACE Calgary Section.

Their support helped make the session possible, and we are grateful for their continued engagement with the broader engineering and project management community.

Graduate Student Opportunities in Generative AI

We are also pleased to share a new opportunity for prospective graduate students. We are inviting applications for Master’s and PhD positions connected to an emerging research program on Generative AI in construction management, with a particular focus on schedule development and related decision-support challenges.

We are looking for candidates with:

• Strong academic background in engineering, applied mathematics, computer science, or related disciplines

• Experience or interest in AI-based scheduling and construction planning

• Experience or interest in fine-tuning, domain adaptation, and domain knowledge enhancement for generative AI models.

Click here to get to the online form, remember to send your resume and CV through by email as instructed on the form:

Online Application Form

Please share this opportunities with interested students within your network.

A shortlist of candidates will be interviewed by our team, together with our research partners, through at least two rounds of interviews.

Upcoming Department of Civil Engineering Event

Finally, we would love you to join an upcoming event hosted by the Department of Civil Engineering at the University of Calgary: Inspiring Voices: Women in Civil Engineering Leadership Dinner.

This will be taking place on Thursday, April 9, 2026 at 6:00 PM at the Last Defense Lounge (LDL) at the University of Calgary Main Campus.

See Event Webpage and Register

The program includes two inspiring featured talks that will bring both depth and personal reflection to the evening. Jess Theroux’s talk, Taking My Time, An Act of Rebellion, invites reflection on resilience, intention, and the power of moving through life and work at a pace that is thoughtful and self-defined.

Later, Jessica Bekker will share Personal Journey in Learning and Discovering the Blackfoot Tipi as an Indigenous Woman and Engineer, offering a deeply personal perspective on culture, identity, learning, and engineering.

Together, these talks promise to make the evening not only engaging, but also meaningful and memorable.

Looking Ahead

This has been a busy and productive year for us as we continue building the seeds of several new research initiatives. We are planning to share more of our latest research in the months ahead. Please stay tuned, and do let us know what you think. We would be very happy to hear from you and to explore how we might work together.

Subscribe now

On Thursday, March 12, 2026, renowned metallurgical engineer Dr. Terence P. (“Terry”) McNulty will present a lecture titled “25 Years of Commissioning & Ramp-Up Performance: McNulty Curves Revisited.”

• Speaker: Dr. Terence P. McNulty

• Date: Thursday, March 12, 2026

• Time: 4:30 PM (MDT)

• Location: University of Calgary, N.W. Campus (Please register for more info)

• Registration is available here.

Only a few spots remain, so we encourage interested participants to register soon.

Why Ramp-Up Performance Matters

One of the least understood, but most critical, phases of major capital projects occurs after construction is completed: the period when facilities transition from commissioning to stable operations.

Across the mining, energy, and industrial sectors, many projects struggle during this stage. Even when construction is completed successfully, achieving design throughput and stable performance often takes far longer than expected. In some cases, it can take months or even years before operations stabilize.

Dr. McNulty’s well-known McNulty Curves offer one of the most widely referenced empirical frameworks for understanding this phenomenon. Developed through decades of case studies across the minerals and process industries, the curves illustrate how few projects achieve full capacity quickly and how varied ramp-up trajectories can be across facilities.

In this lecture, Dr. McNulty will revisit and update the curves based on additional case histories and share insights on:

• Why many projects miss early throughput targets

• The operational and managerial factors that influence ramp-up success

• What distinguishes fast, stable start-ups from prolonged underperformance

• Practical strategies project teams can use to reduce start-up risk

For professionals involved in project management, commissioning, operations readiness, and asset performance, these lessons are particularly valuable.

Learning from a Pioneer in the Industry

Dr. McNulty brings more than sixty years of experience across research, operations, and executive leadership in the minerals industry.

Over his career he has worked with organizations such as Anaconda Company, Kerr-McGee Chemical, and Hazen Research, and later founded T.P. McNulty and Associates, a global consulting firm specializing in metallurgy, mining, and process engineering.

His work has influenced generations of engineers and project managers seeking to understand the operational realities of large industrial facilities. With more than 50 publications, multiple patents in copper metallurgy, and recognition from leading professional societies, Dr. McNulty’s insights combine rigorous technical understanding with extensive real-world experience.

Event Sponsors

This event has been made possible through the generous support of:

• Canadian Institute of Mining, Metallurgy & Petroleum (CIM)

• EPM Research Group at the University of Calgary.

• Hatch

• AACE Calgary Section

Takeaways,

• Risk management failures are often rooted in faulty communication.

• Ask AI to produce multiple versions and choose one deliberately.

• Value clarity, define glossaries, and lock the semantics.

• Maintain AI-generated risk registers separately with clear rules for escalation.

• Train young engineers to challenge AI outputs.

Artificial intelligence is flooding into project management tools, knowledge bases and decision support systems.

For engineers and project leaders it is easy to see the appeal: a tireless assistant that drafts schedules, writes progress reports and crunches massive data sets in seconds.

However, the very ease with which AI is being woven into everyday workflows should give risk professionals pause.

With any powerful technology, the earliest waves of adoption are the most risky. Initial successes can create a sense of confidence long before underlying vulnerabilities have been truly tested.

This emerging landscape is increasingly showing a familiar pattern: over-trusted outputs, underspecified models, and opaque data pipelines that quietly compound into poor decisions, or, at times, significant failures.

At the EPM Research Lab at the University of Calgary, our research focuses on learning the challenges and uncertainties around AI adoption from our from industry partners. Drawing on these shared lessons, I aim to outline several practical ideas and the critical issues teams should watch for when integrating AI into project and risk management workflows.

Communication Failures in Risk Management

Human communication sits at the core of project and risk management. When it fails, the consequences can be severe.

Many well-known engineering disasters were not rooted solely in technical errors, but in misunderstandings, ambiguous language, and misaligned interpretations. This is the lens through which the adoption of AI in risk management must be viewed.

For those of us in Calgary, a striking example of how miscommunication can trigger major project challenges lies at the very heart of our city’s most recognized landmarks.

Calgary’s Peace Bridge

Calgary’s Peace Bridge highlights how semantic and standards misalignment, not a purely technical defect, can derail a major project.

The welds fabricated in Spain met European standards, yet failed under Canadian testing protocols. The project relied on an unverified assumption that international specifications were interchangeable. That gap in interpretation became a critical interface risk.

When Canadian inspectors applied domestic code-based probe testing, dozens of welds were deemed non-compliant, prompting extensive re-inspection, grinding, and re-welding of the tubular steel structure [1].

The rework increased the total weld length from roughly 500,000 inches to over 1.5 million inches, requiring months of additional labour and specialist oversight, effectively tripling the welding effort. The opening date slipped repeatedly, from an initial 2010 target to early 2011, then mid-2011, and ultimately to March 2012, demonstrating how a single unmanaged interface risk between design codes can cascade into substantial schedule overrun [1].

“In a timely fashion”

The 2021 collapse of the Champlain Towers South condominium in Surfside, Florida illustrates how ambiguous technical language can become a fatal failures.

A 2018 engineering report warned of “major structural damage” and advised that deterioration be repaired “in a timely fashion.” In hindsight these phrases were dire, but non-engineers on the condo board struggled to interpret how “major” the damage was or what timeframe “timely” implied.

Crucially, the report never stated explicitly that the structure was at risk of collapse. The intended urgency did not translate, decisions were delayed, and 98 people died when the building partially collapsed [3].

Other well-known disasters show similar patterns. The 1986 Challenger shuttle explosion has been described as a technical communication failure: engineers raised concerns about O-ring performance at low temperatures, but these concerns were not forcefully communicated into the final launch decision [4]. Communication breakdown also played a central role in the Deepwater Horizon blowout, where early warning signs and technical concerns were not effectively escalated in real time [5]. The Hyatt Regency walkway collapse resulted from a critical design change that was misunderstood between the steel fabricator and the engineering firm [6]. More recently, Boeing’s 737 MAX crashes exposed how software dependencies, documentation, and training gaps can create semantic blind spots between designers, regulators, and pilots [7].

Across these examples, the pattern is consistent and reflects the main lens through which the adoption of generative AI in risk management must be viewed.

Avoid Anchoring and Biases to AI

Behavioral science has shown that human judgments are systematically biased [8].

In a classic experiment, two groups of participants were asked to estimate the product of a sequence of numbers. When the sequence started with larger numbers (8 × 7 × 6…), median estimates were far higher than when the sequence started with smaller numbers. The early figures anchored people’s judgments [9].

In project planning, anchoring occurs when stakeholders fixate on original schedules or budgets and resist adjustment even as new information emerges [8,9]. In risk management, AI-generated risk descriptions can create a similar effect by anchoring how teams interpret and understand a risk.

Generative AI systems exacerbate anchoring because their outputs feel precise. Under the hood, large language models (LLMs) are stochastic: each response is one sample from a probability distribution over possible completions. Yet the language is fluent and confident, which tempts teams to treat the first answer as “the answer.”

Aside from anchoring, the NIST Generative AI Risk Management Framework highlights how human-AI configurations can create automation bias and over-reliance [10]. Users may accept AI outputs even when there are warning signs, particularly when they do not understand the model’s limitations.

Over-reliance also manifests as sycophancy bias: chatbots that align with user cues, echoing opinions to please or reassure the user. Chatbots affirm users’ actions more often than human interlocutors and rarely challenge harmful framings, thereby reinforcing existing beliefs rather than interrogating them [11].

At the same time, evaluations of AI-generated summaries show that models frequently omit qualifiers and over-generalise findings. When explicitly instructed to “be accurate,” some models paradoxically increased the rate of over-generalisation [12].

Self‑consistency and best‑of‑N sampling

AI research has started to address these issues through multiple-output decoding strategies such as self-consistency and best-of-N sampling.

The idea of self-consistency is proposed to generate several independent reasoning paths (using chain-of-thought prompting) and then selects the answer that is most consistent across those paths [13]. Best-of-N sampling extends this idea: the model generates multiple candidate outputs and then selects the one with the highest internal confidence or best fit to predefined criteria [14].

Both strategies reduce hallucinations and improve reasoning by averaging across several “runs” rather than trusting a single sample. Practitioners can adopt a simple analogue of these techniques:

For project teams, treat AI outputs like Monte Carlo draws. Request several versions, review the spread and pick the one that best fits your domain knowledge. This simple practice not only mitigates anchoring but also surfaces oversights and faulty assumptions.

The Rising Cost of Ambiguity

The growing reliance on generative AI introduces a new and more pervasive form of under-specification.

Because LLMs have reduced the effort of producing confident, coherent language, project teams now face an environment saturated with plausible explanations and risk narratives that differ only subtly in meaning.

A study of 14 widely used LLMs found substantial intra-model variability: the same LLM, with the same prompt, may produce outputs ranging from poor to highly creative [15].

In this context, any ambiguity in the definition of terms creates space for the model to silently reinterpret them, and as the volume of AI-generated text expands, the organizational cost of semantic ambiguity rises proportionally.

This dynamic is particularly evident in LLM-assisted workflows, where under-specification manifests as semantic drift. When prompts are loosely defined, models fill the gaps with their own latent assumptions, producing multiple internally coherent but conceptually incompatible outputs [16].

The challenge going forward will be ensuring that the documentation reflects a stable, shared interpretation. The only antidote is investing in clarity: explicitly defining ontologies, articulating precise thresholds, constraining prompts, and requiring structured outputs that force the model to operate within the organization’s established vocabulary.

In conventional machine learning pipelines, under-specification arises because many different parameter configurations can achieve similar performance on the training and validation distributions while behaving differently under new conditions [17].

LLMs reproduce a similar phenomenon, but through language rather than model parameters. Both cases highlight the same lesson for project management: without rigor in definitions, structure, and semantic alignment, AI-enabled workflows risk divergent interpretations into the decision-making.

Risk Identification Paradox

Construction projects have long aspired to “thorough” risk identification, yet practitioners face a persistent paradox: the more exhaustive the pursuit of risks becomes, the less effective it often is [18].

Excessively long risk registers obscure underlying patterns, dilute managerial attention, and overwhelm already stretched delivery teams, allowing weak signals of consequential risks to disappear within a series of low-value items [19].

The opportunity cost is substantial. Every hour spent documenting remote hypotheticals is an hour not invested in analyzing, mitigating, or rehearsing responses to high-impact threats. High-reliability sectors have long recognized that there is a rational stopping point.

The ALARP principle (As Low As Reasonably Practicable) operationalizes this idea by acknowledging that once risks fall below a tolerability threshold, additional mitigation effort offers diminishing returns [18]. The Pareto principle offers a similar insight: a small subset of risks typically accounts for most of the potential impact.

Modern enterprise-risk frameworks embed these concepts directly. COSO’s ERM framework and ISO 31000 address risk appetite and tolerance explicitly, giving organizations structured cut-off points for attention and action [21,22].

When combined with the reality that projects unfold in dynamic and uncertain environments, best practice reframes risk identification as an iterative, context-responsive activity, not a one-time attempt at omniscience.

This highlights the communication dimension within the broader risk-management process: risk registers are, above all, communication instruments. Their purpose is to keep teams collectively aware of emerging issues, coordinate controls, and maintain shared situational awareness [23].

Tiered or multi-level risk registers reinforce this role by distinguishing between material or urgent risks that require leadership attention and other less crucial risks that can be monitored locally.

Escalation processes then serve two complementary functions: signaling (alerting leaders to cross-boundary or systemic issues) and screening (filtering noise before it reaches governance forums). When designed well, these architectures help maintain organizational mindfulness without overwhelming project teams or steering committees [24].

AI-Generated Exhaustive Registers

Generative AI and large language models now make it technically feasible to produce highly exhaustive, dynamic, and continuously refreshed risk registers.

These systems can ingest real-time project data, design revisions, schedule updates, and change requests, generating a broad constellation of potential risks far faster than human teams.

Early research indicates that LLMs can outperform human groups in the breadth of risk identification, provided their outputs are curated [25]. Recent studies show that LLMs can surface rich risk sets for sustainable operations in onshore wind projects, again with the caveat that human filtering remains essential [26].

In this sense, AI is emerging as a high-recall “risk discovery engine,” particularly valuable in megaprojects where interfaces and stakeholder dynamics evolve rapidly.

However, exhaustiveness cannot come at the expense of human judgment or efficiency. Behavioral research demonstrates that when information volumes grow without structure, individuals revert to heuristics or defer to seemingly authoritative systems [8,17,23].

A more practical architecture, therefore, is to maintain a separate AI-generated risk register that operates in parallel to the formal project register.

The AI tiered system can continually populate this space with emerging signals, weak indicators, and speculative risks without burdening the project team. From there, organizations should define explicit escalation criteria.

Only those AI-generated risks that meet the agreed criteria are then elevated into the formal project risk register, where they receive structured analysis, ownership, and monitoring.

This approach preserves the advantages of broad, AI-driven exploration while safeguarding the constraints of human attention and decision-making. It mirrors principles from resilience engineering: sense widely at the periphery, where anomalies first emerge, but escalate selectively at the core, where decisions and resources must remain focused [18,23].

Human Oversight Means Training People to Challenge AI

Many experienced professionals can readily distinguish between plausible AI reasoning and obvious AI errors.

But the next generation of engineers may not be able to do so without proper mentoring and support. It is not guaranteed that this intuition will persist without explicit preparation [27].

As AI-generated text increasingly permeates the project environment, risk registers, engineering memos, meeting minutes, it will become harder to spot flawed logic embedded within these artefacts, especially for future professionals who have never performed these tasks without AI assistance.

For this reason, simply stating that “humans are in the loop” is insufficient. The NIST Generative AI Profile emphasizes that human oversight can fail when users become over-confident in AI systems or when organizational routines implicitly defer to automated outputs [10]. Similarly, a recent CSET report on automation bias documents how over-reliance on automated systems has contributed to accidents and near-misses, underscoring that effective oversight must be deliberately designed and continually trained [28].

The industry must therefore invest in cultivating AI literacy among young engineers, planners, and inspectors. This includes embedding training on critical thinking, ethics, and model limitations into graduate and early-career programmes, and creating mentorship structures where experienced project managers actively explain why certain AI-generated recommendations are implausible or unsafe. Put plainly, organizations must hire and train young engineers to challenge AI outputs, not defer to them.

Question then Becomes

As generative AI embeds itself deeper into engineering and project management, the central challenge increasingly shifts away from computation and towards communication.

To navigate this era safely, the construction industry should treat AI as an accelerator of both strengths and vulnerabilities.

Organizations that invest now in semantic discipline, reliable AI workflows, and training young engineers to interrogate AI will be better positioned than those that simply add ‘AI’ to their tool stack.

For decades, the saying in construction disputes has been that “the party with more documentation wins.” But as AI becomes ubiquitous, and every interaction, revision, and decision is automatically captured, will the advantage shift to the party with clearer semantics and shared understanding?

As the generation of engineers who can challenge AI outputs ages and retires, will the next generation, raised on AI-assisted reasoning, retain the ability to question, verify, and audit these systems?[27] Or is the trajectory more likely to get worse before it gets even worse?

And as AI assumes more of the administrative load in project and risk management, will it free practitioners to focus on the human communication that actually mitigates risk? Or will it widen the gap between documentation and understanding?

Notes

Join the EPM Network to access insights, influence our research, and connect with a community shaping the industry’s future.

Subscribe now

Support us by sharing this article with your friends and colleagues, or over social media.

If you wish to share your opinion, provide insights, correct any details in this article, or if you have any questions, please email editor@epmresearch.com.

Refer to this article using the following citation format:

• Zangeneh, P. (2025), “Industry Should Brace for the First Wave of AI-Driven Risk Management Failures ”, EPM Research Letters.

References

1. Journal of Commerce. Welding problems cause Calgary Peace Bridge delays. Journal of Commerce. 2011 Apr 6.

2. CBC. Welding issues delay Peace Bridge project. 2010 Nov 16.

3. Forbes. When words kill: lessons from the Champlain Towers collapse. Forbes. 2021 Jun 28.

4. NASA. Lessons From Challenger. 2021.

5. De Wolf, Daniel. Crisis communication failures: The BP case study. International Journal of Advances in Management and Economics. 2013.

6. National Academy of Engineering. The Hyatt Regency walkway collapse. Online Ethics Center.

7. Herkert, J., Borenstein, J. & Miller, K. The Boeing 737 MAX: Lessons for Engineering Ethics. Sci Eng Ethics.

8. Kahneman, Daniel. Thinking, fast and slow. macmillan, 2011.

9. Decision Lab. Anchoring bias: the psychology of clinging to initial estimates. The Decision Lab. 2023.

10. National Institute of Standards and Technology (NIST). Artificial Intelligence Risk Management Framework: Generative AI Profile (NIST AI 600-1). 2024.

11. Hern A. Sycophantic AI chatbots tell users what they want to hear, research finds. The Guardian. 2025.

12. Peters U, Chin-Yee B. Generalization bias in large language model summarization of scientific research. R. Soc. Open Sci. 2025.

13. Wang X, Wei J, Scales N, et al. Self-consistency improves chain-of-thought reasoning in language models. In: Proceedings of ICLR 2023 [Internet].

14. Kang, Zhewei, Xuandong Zhao, and Dawn Song. Scalable best-of-n selection for large language models via self-certainty. arXiv. 2025.

15. Haase, Jennifer, Paul HP Hanel, and Sebastian Pokutta. “Has the Creativity of Large-Language Models peaked? An analysis of inter-and intra-LLM variability.” Journal of Creativity. 2025.

16. Yang, Chenyang, et al. What Prompts Don’t Say: Understanding and Managing Underspecification in LLM Prompts. arXiv. 2025.

17. D’Amour A, Heller K, Moldovan D, et al. Underspecification presents challenges for credibility in modern machine learning. J Mach Learn Res. 2022.

18. Fenton N, Neil M. Risk assessment and decision analysis with Bayesian networks. 2nd ed. Boca Raton: Chapman & Hall/CRC; 2019.

19. Turner BA, Pidgeon NF. Man-made Disasters. 2nd ed. Oxford: Butterworth-Heinemann; 1997.

20. Health and Safety Executive (HSE). Reducing risks, protecting people: HSE’s decision-making process (R2P2). HSE. 2001.

21. Committee of Sponsoring Organizations of the Treadway Commission (COSO). Enterprise Risk Management—Integrating with Strategy and Performance. COSO. 2017.

22. International Organization for Standardization. ISO 31000:2018 – Risk management: Guidelines. ISO . 2018.

23. Weick KE, Sutcliffe KM. Managing the unexpected: resilient performance in an age of uncertainty. 2nd ed. San Francisco: Jossey-Bass; 2007.

24. Zangeneh, P. (2025), The Art of Strategy - Part 2: A Review of Dixit and Nalebuff’s Classic Within the Context of Project Management, EPM Research Letters.

25. Nyqvist M, Landberg M, Josephson PE. Can ChatGPT exceed humans in construction project risk management?. Safety Insights. 2024.

26. Wen H, AbouRizk S, Mohamed Y. Using large language models to identify project risks for sustainable operations: a case study of onshore wind farms. In: CIB Conferences. 2025.

27. Cosentino R. Navigating Major Programmes podcast: human-specific skills vs AI in major project controls. 2025.

28. Center for Security and Emerging Technology (CSET). AI safety and automation bias. CSET. 2023.

This year, I have the privilege of co-chairing the conference with Dr. Fuzhan Nasiri from Concordia University. Together, and with the remarkable support of our organizing committee, we are shaping this specialty track into a vibrant meeting ground for both academics and practitioners who see engineering management not just as an analytical discipline but as a driver of real-world transformation.

I wish to encourage you to participate in this important Canadian event, and hope to see you and the broader EPM Research community in Québec City next June!

Submit your abstract by October 22, 2025, and join us in sharing your ideas, and shaping the future of engineering management in Canada.

Submit Your Abstract

Call for Abstracts: Now Extended

I’m happy to learn that the local organizing committee has extended the deadline for abstract submissions to October 22, 2025.

We invite contributions in three flexible formats of full papers, extended abstracts, and case studies. Each format serves a distinct purpose in advancing the conversation between research and practice:

• Full Papers – For authors aiming to present rigorous research, analytical frameworks, or methodological advancements. These contributions typically include data, models, and detailed findings that push the theoretical or technical boundaries of engineering management.

• Case Studies – For practitioners and researchers seeking to share lessons from real projects. These submissions highlight decisions, trade-offs, and reflections from implementation, successes, failures, or unexpected outcomes that can inform future practice.

• Extended Abstracts – Especially well-suited for industry participants. This format allows professionals to discuss emerging issues, operational challenges, and innovation ideas without the need to develop a full paper. Extended abstracts encourage early sharing of insights, enabling dialogue around problems that merit joint academic-industry attention.

Themes for 2026

This year’s topics reflect the evolving frontiers of engineering management in Canada and beyond that was put together by the our great members of the engineering management technical committee:

1. Management of Generative AI in Civil Engineering Applications

2. Project Delivery, Contracts, and Collaborative Models

3. Civil Engineering Workforce and Trades Management

4. Public Policy and Civil Engineering in Public Sector

5. Innovation, Technology Management, and New Business Models in Civil Engineering

6. Procurements and Supply Chain Management

7. Risk, Uncertainty, and Complexity Management and Modelling

8. Leadership and Stakeholders Management in Civil Engineering

9. Buildings, Facilities, and Infrastructure Management

10. Construction and Project Management

11. Management of Climate Change Adaptation & Resilience Projects

Submit your abstract by October 22, 2025, and join us in sharing your ideas, and shaping the future of engineering management in Canada.

Submit Your Abstract

Building a Stronger Dialogue with Industry

One of our major efforts this year is to expand participation from Canadian public and private industry. With the help of our committee members, we are reaching out to project owners, utilities, contractors, and consultancies who make engineering decisions at scale.

We aim to make this conference a place where research meets real constraints, and where industry voices help shape the next generation of tools, methods, and engineering leaders of Canada.

Looking Ahead: Bi-Weekly Articles on EPM Research

Starting this November, our academic year series will start with bi-weekly updates featuring short project case studies and curated articles on emerging topics in project and engineering management.

Subscribe now

This review draws on key concepts from game theory and negotiation literature to support better decision-making in project settings. Here we review:

1. Thomas Schelling and Strategic Moves

2. Conflicts as Bargaining in Variable-sum Games

3. Deterrence

4. Rational Irrationality

5. Bargaining

6. Negotiations

7. Credibility of Strategic Moves in Project Environments

8. Question then Becomes

I first revisit the core concepts of strategic moves and their critical components: commitments, threats, promises, and credibility from the reviews we did of the Art of Strategy by Dixit and Nalebuff (Part 1 and Part 2) (1).

Then I explore how these principles apply to project environments, and delve into broader idea of conflict deterrence, and limiting rationality, first formalized by Thomas Schelling in his seminal book The Strategy of Conflict that dates back to 1960 (2).

The significance of Schelling’s work lies in how it lays the foundation for a strategic approach to conflict, bargaining, and negotiation. Despite the seemingly cold-blooded title of his book, Schelling clarifies that the theory is not about the efficient application of violence. It is not fundamentally a theory of aggression, resistance, or war. Rather, it is a theory of interdependent decisions and nonuse of violence and the conditioning of one’s behavior based on the behavior of others.

We then review another seminal work, Getting to Yes: Negotiating Agreement Without Giving In by Roger Fisher, William Ury, and Bruce Patton, to discuss core concepts such as the Best Alternative to a Negotiated Agreement (BATNA) and its critical role in negotiation strategy (3).

As Schelling points out, the strategy of conflict applies across many areas of daily life. Whether dealing with children, navigating right-of-way in traffic, or interacting with friends and partners, most of our social behavior involves elements of conflict, bargaining, and negotiation.

Thomas Schelling and Strategic Moves

Schelling was a pioneer in exploring the idea of not merely playing the game but changing the game through strategic moves.

As discussed in the first and second parts of the review of the Art of Strategy by Dixit and Nalebuff, strategic moves involve actively shaping the game itself to influence outcomes and the behavior of others. These moves can be categorized into three core types: commitments, threats, and promises.

Strategic moves consist of two critical components: the move itself and the course of action that establishes its credibility. The effectiveness of any strategic move hinges on both the action and the credibility behind it.

Commitments are unconditional actions that restrict one’s own choices to compel a response. They are powerful tools that can bind opponents to a specific course of action. However, undermining an opponent’s commitment—if done strategically—can open pathways for de-escalation and negotiation.

Hernán Cortés famously "burned his ships" during the 1519 conquest of the Aztec Empire. While the historical accuracy of this account is debated, it serves as a powerful metaphor for total commitment—eliminating the option of retreat. On the flip side, as Sun Tzu noted, undermining an opponent’s commitment may require offering them—if not a literal one—a golden bridge to retreat across.

In contrast, threats and promises are conditional moves with distinct dynamics. Threats aim to deter undesirable actions by imposing negative consequences. They are typically open-ended and often seek to preserve the status quo. Promises, on the other hand, incentivize desired actions through rewards and are usually intended to disrupt or improve the status quo.

Conflicts as Bargaining in Variable-sum Games

Schelling starts his book by establishing that pure conflict—where interests are entirely opposed—is an exceptional case. It arises only in scenarios of total destruction, a rarity even in warfare. More often, conflict involves a complex interplay of competing interests, where success is not defined by absolute victory over an adversary but by achieving favorable outcomes within one’s own value system.

In this context, "winning" means securing gains that align with one's goals and priorities rather than defeating the opponent outright. This can occur through bargaining, compromise, or by avoiding mutually destructive actions altogether.

Conflict management, therefore, is not solely about overpowering an opponent but about skillfully navigating the potential for conflict. Strategies such as deterrence, limited engagement, and negotiation highlight the importance of mutual dependence and shared interests. Whether in warfare, business negotiations, or labor strikes, the objective is often to find solutions that prevent costly and damaging confrontations.

Most conflict situations resemble bargaining scenarios more than zero-sum contests. Unlike constant-sum games, where one party’s gain equates to another’s loss, many conflicts are variable-sum games in which mutually beneficial outcomes are possible.

This perspective underscores the importance of understanding the other party’s motivations and decisions, as success often hinges on finding common ground rather than imposing unilateral solutions.

For example, a successful labor strike is not necessarily one that devastates the employer; it may be one that is never happened. As Sun Tzu stated, "The greatest victory is that which requires no battle."

Deterrence

Framing conflict as an essential type of bargaining, the concept of Deterrence then hinges on the credible threat of force, not its actual use.

It is the art of persuading an adversary that certain actions will incur unacceptable costs, not just to the adversary but to those making the threat too. For this to work, the threat must be believable, which depends on the perceived willingness and capability of the deterring party to follow through.

A key distinction lies between using force and exploiting its potential. Effective deterrence demands more than military strength, it requires strategic communication, psychological insight, and an understanding of the adversary’s calculus.

Deterrence operates where conflict and common interest coexist. It fails in cases of complete alignment or irreconcilable hostility. This dynamic applies not just to adversaries but also to allies, consider how nations deter partners from defecting on agreements, much like a state might deter an enemy.

Deterrence becomes critical in settings where no higher authority exists to enforce rules. Ancient feuds and modern criminal networks illustrate this dynamic: gangs rely on threats—not courts—to uphold agreements. Often limited conflict, retaliation, and fragile truces emerge as substitutes for formal enforcement among nations.

Rational Irrationality

Rationality, as a guiding principle in decision-making, often implies logical consistency, optimization of outcomes, and alignment with defined objectives. However, in conflict scenarios, this notion of rationality is not universally advantageous.

A purely rational actor is predictable, calculable, and sometimes exploitable. This creates a paradox: in certain strategic interactions, it may be advantageous to *limit* one’s own rationality.

Strategic interactions often reveal that strict adherence to rational decision-making can be a disadvantage. In certain contexts, especially where unpredictability or deterrence is required, the appearance of irrationality can be a valuable asset.

Rationality encompasses not only cognitive processes but also the systems, structures, and agents through which decisions are enacted. These include technological tools, communication channels, legal frameworks, and even the rationality of one’s collaborators. From a strategic standpoint, disrupting or limiting these rational components may be beneficial. For example, impairing communication systems or legally constraining access to resources may serve as deliberate tactics to counter external coercion or exploitation.

Bargaining

Most real-world conflicts resemble variable-sum games where mutual gains are possible. Yet, achieving favorable outcomes is not only a matter of efficiency but also of distributional strategy—determining who gains more from the agreement.

The conventional wisdom that effective negotiators should have full authority and flexibility to make concessions is misleading. Often, the power of a negotiator lies precisely in their apparent inability to yield or compromise. A negotiator who can credibly claim to lack authority to make concessions possesses a strategic advantage. By limiting their own flexibility, they force the other party to adjust its position.

Schelling’s insights into commitment strategies highlight how deliberately restricting one’s options may enhance bargaining power.

For commitments to be effective, they must be credible, visible, and costly to break. The mechanisms that enforce these commitments range from formal contracts and technological solutions to cultural norms and reputational stakes. In certain contexts, verbal oaths or public statements may suffice; in others, formalized agreements or technological tools are required to ensure credibility. The effectiveness of these commitments often hinges on institutional enforcement or social expectations that render renegotiation prohibitively costly.

The strategic use of commitments also involves timing and communication. In scenarios where commitments are made sequentially, the first mover often holds a decisive advantage by setting terms that subsequent actors must either accept or challenge under constrained conditions.

However, when communication is delayed or obstructed, as in negotiations conducted through indirect channels, parties with limited receptiveness to counteroffers may inadvertently strengthen their bargaining position. By appearing or genuinely being unable to reconsider their stance, they create an impression of unyielding resolve.

Ultimately, Schelling posits that the paradox of bargaining power lies in the deliberate cultivation of weakness. By surrendering certain options or reducing one’s flexibility, a negotiator can alter the expectations and strategies of their opponent. This counterintuitive principle applies across a wide spectrum of bargaining scenarios, from labor disputes and business negotiations to geopolitical standoffs.

Negotiations

The concept of BATNA—Best Alternative To a Negotiated Agreement—was developed by Roger Fisher and William Ury in their book Getting to Yes: Negotiating Agreement Without Giving In.

Negotiations are seldom balanced; differences in power—whether financial resources, political connections, or organizational influence—can significantly impact outcomes. While no negotiation method can guarantee success against a stronger opponent, strategic preparation can greatly enhance your position. The core strategy here is your BATNA.

A common error is to rely on a rigid "bottom line"—a fixed threshold beyond which an offer will not be accepted. While bottom lines can protect negotiators from poor deals, they also severely limit flexibility. Such rigidity prevents negotiators from identifying innovative solutions that better address underlying interests.

In contrast, clearly defining your BATNA enables more strategic and dynamic negotiation. A BATNA represents your most favorable option if negotiations fail. Instead of setting a rigid boundary, you measure proposed agreements against this best alternative. This approach safeguards against overly optimistic or pessimistic expectations about potential outcomes.

To illustrate, consider a family selling their home. Rather than simply stating a minimum acceptable price, they should explore their BATNA: Would renting the property or delaying the sale be preferable to accepting a lower offer? Identifying and clearly understanding their BATNA empowers them to evaluate offers effectively and avoid agreements that are detrimental to their interests.

Effectively using BATNA requires intentional development through three critical steps:

1. Identify Alternatives: Generate a comprehensive list of possible actions if negotiation fails.

2. Improve Feasibility: Refine the most promising alternatives into actionable, realistic options.

3. Select the Best Alternative: Choose the strongest alternative to serve as your benchmark during negotiations.

A robust BATNA strengthens your negotiating power. Consider job interviews: entering negotiations with multiple job offers significantly shifts the power dynamic in your favor compared to having none.

The strategic disclosure of your BATNA should be carefully evaluated. Revealing a strong BATNA can enhance your leverage; however, disclosing a weak BATNA can weaken your negotiating position. Similarly, understanding the other party's BATNA provides crucial insight. If their alternative is overly optimistic, helping them assess it more realistically can create opportunities for mutually beneficial agreements.

The BATNA framework is essential in project management contexts. Leaders who rigorously analyze and strengthen their BATNA position their teams effectively, mitigating vulnerabilities during negotiations and optimizing outcomes. Whether negotiating contracts, partnerships, or internal resource allocation, clearly defining and continuously improving your BATNA is a cornerstone of strategic negotiation.

Credibility of Strategic Moves in Project Environments

Dixit and Nalebuff classified the approaches to establishing credibility for strategic moves into four categories of tactics:

1. Changing Payoffs: Restructuring the game to make compliance more attractive or deviation prohibitively costly.

2. Restricting Choices: Limiting options to reverse commitments, effectively creating credible threats or promises.

3. Incremental Moves: Using repeated, smaller engagements to build trust incrementally.

4. Using Agents and Teams: Employing external entities or teams to enhance credibility and accountability.

Among these tactics, restricting choices through automated systems, contractual terms, or similar measures is the most readily available and practical option for project managers. In contrast, tactics such as employing third-party arbitrators or auditors (agents and teams), significantly changing payoff structures, or incorporating incremental moves usually require higher levels of arrangements than of the daily plans of project managers, making them less immediately accessible.

Projects, by definition, are temporary endeavors aimed at achieving specific, unique outcomes within clearly defined timeframes and resources. Unlike ongoing business operations, project settings often lack the luxury of repeated interactions between the same stakeholders, fundamentally altering how credibility and trust are established.

While reputation within the project-based industry is critical—given the interconnected nature of project stakeholders and the limited pool of clients, contractors, and consultants—the traditional tactic of incremental moves, which relies heavily on repeated engagements to foster trust, is typically less accessible.

Nonetheless, contracting strategies such as progressive contracting are essentially aiming to bridge this gap. Progressive contracting involves dividing the overall scope into sequential phases or milestones, each with clearly defined deliverables and decision points. At each stage, parties assess performance and decide whether and how to proceed.

Question then Becomes

As this review has shown, managing conflict in project environments requires more than interpersonal skills. It demands a strategic mindset rooted in an understanding of bargaining, credibility, and negotiation dynamics.

Drawing from Schelling’s work on strategic moves and deterrence, and complemented by Fisher and Ury’s practical insights on BATNA, we see that conflicts in projects are rarely zero-sum. They more often resemble variable-sum games, where value can be created or lost depending on how threats, promises, and commitments are made and enforced.

Yet applying these principles in real project settings raises important questions. For instance: How can project managers strategically limit their own flexibility to strengthen their bargaining position without also giving in to the paradox of bargaining power and appearing weak?

In what ways and to what extent does the absence of a central enforcement authority manifest in multi-stakeholder megaprojects, particularly between contractors? Does the presence of a client always serve as an effective arbiter, or are there conditions where deterrence becomes necessary?

Notes

Join the EPM Network to access insights, influence our research, and connect with a community shaping the industry's future.

Subscribe now

Support us by sharing this article with your friends and colleagues, or over social media.

If you wish to share your opinion, provide insights, correct any details in this article, or if you have any questions, please email editor@epmresearch.com.

Refer to this article using the following citation format:

• Zangeneh, P. (2025), “Conflict, Bargaining, and Negotiation in Project Management - Exploring Core Concepts and Thomas Schelling’s Insights in Project Management Context.” EPM Research Letters.

This article is a reading material for the “Strategy in Projects: Game Theoretic Approach” course at the University of Calgary, Department of Civil Engineering.

References

• Dixit, A. K., & Nalebuff, B. J. (2008). The art of strategy: A game theorist's guide to success in business and life. W. W. Norton & Company.

• Schelling, T. C. (1980). The strategy of conflict (2nd ed.). Harvard University Press. (Original work published 1960)

• Fisher, R., Ury, W., & Patton, B. (2011). Getting to yes: Negotiating agreement without giving in (3rd ed.). Penguin Books. (Original work published 1981)

It was his obsession, however, that nudged open the door to one of the universe’s most profound laws of reasoning.

What began as a quiet attempt to reason from effect back to cause would, centuries later, evolve into an entire framework for understanding complexity, risk, AI, and adaptive decision-making.

In this article, we begin with the early history of Bayes' theorem and trace its evolution to modern-day applications in risk modeling. We review:

1. What was wrong with Reverend Bayes?

2. Bayes’ Pool Table Thought Experiment

3. Mathematics of Risk

4. From Thomas Bayes to Bayesian Networks

5. Three Fundamental Patterns of Probabilistic Interdependence

6. Risk Modeling with Bayesian Networks

7. Bayesian Networks in Projects and Operations

8. Bayesian Networks in Digital Twins

9. Question then Becomes

Bayesian Networks (BNs), build on Bayes’ original insight by modeling the probabilistic dependencies between variables using directed acyclic graphs (DAGs). These networks capture how real-world phenomena influence one another, not through isolated causes, but through structured chains of probabilistic dependencies (1–4).

Over the years, Bayesian methods have proven to be powerful tools for tackling real-world challenges in projects and operations risk modeling, and systems engineering. This article is part of a course developed at the University of Calgary, focused on practical applications of Bayesian methods.

What was Wrong with Reverend Bayes?

From limited information we have, it appears that Reverend Thomas Bayes (1702–1761) was a quiet intellectual, driven by both religious faith and a rigorous sense of logic.

In his later years, he turned to a deceptively simple question: How should we update our beliefs in light of new information? Perhaps he recognized that the human brain does this instinctively, given wet sidewalks, we infer rain. But his goal was to mathematically determine the likelihood of a hidden cause given an observed effect.

Unlike classical probability, which predicts outcomes based on known conditions, Bayes sought to reverse the lens: to reason backwards from data to underlying causes.

Bayes, however, never articulated his findings using today’s formalism. After his death, Richard Price, his friend and colleague found an unfinished paper in his files titled An Essay Towards Solving a Problem in the Doctrine of Chances (1763). Realizing its importance, Price edited, completed, and prepared the work for publication. He also wrote an introduction to explain and contextualize the significance of Bayes’ result. The torch was later carried forward by Pierre-Simon Laplace, who extended Bayes’ ideas into a broader and more influential framework.

Bayes’ Pool Table Thought Experiment

Imagine a pool table stretching from 0 to 1m. Now let us consider the following three acts.

First - Somewhere along this table, a wooden bar is placed at a secret location, call it X. The exact position is hidden from us; all we know is that it's somewhere between 0 and 1, chosen uniformly at random. This unseen position represents a probability we want to uncover.

Second - A ball is dropped. But this one land according to a rule—it falls randomly somewhere between 0 and X, again chosen uniformly, call it Y.

Third Act – The wooden bar is removed. The only thing we can observe is Y.

And now the essential question arises: given this observed Y, what can we reasonably infer about the unknown X? In other words, how do we revise our belief about where the wooden bar might have been placed?

Engaging with the Experiment

First, the hidden value X—the location of the wooden bar—is assumed to be uniformly distributed between 0 and 1. This means, before observing anything, we treat all positions as equally likely.

Second, the observed value Y, the landing spot of the ball, and we know is also uniformly distributed over the interval between 0 and X. So once X is fixed (though unknown to us), Y is equally likely to land anywhere from 0 up to that value.

When we observe Y, one thing becomes immediately clear: the unknown value X must be greater than Y. The ball couldn’t have landed outside the range from 0 to X, so Y < X.

But beyond that, something more subtle happens. We can reason those smaller values of X—those just slightly larger than Y—are more likely than much larger ones. Why? Because for Y to have landed near itself within a large interval (say from 0 to 0.9), it would have to be a rather rare, although not impossible, outcome.

This asymmetry shifts our belief. After seeing Y, we no longer consider all values of X equally likely. We now prefer those that are closer to Y, while still greater than it. In Bayesian terms, we’ve updated our prior belief (from X uniformly distributed between 0 and 1) into a posterior belief—one that leans toward values of X just above the observed Y.

What we now call Bayes’ rule—or Bayes’ theorem—is a direct consequence of the basic rule of conditional probability that Reverend Thomas Bayes used to calculate the probability of a hidden cause X, given an observed outcome Y, also known as the Posterior distribution of P(X|Y).

In the context of the table and ball experiment, applying Bayes’ rule for values where X > Y leads to a specific form of the posterior distribution that is shown in Figure 2.

For example, let’s say Y = 0.3. Before observing anything, we assume X follows a uniform distribution between 0 and 1—represented as the red line in Figure 1. But after seeing Y = 0.3, the updated belief (the blue curve) tilts in favor of values just above 0.3. As expected, smaller values of X closer to 0.3 are now more probable than those closer to 1.

Importantly, the area under the posterior curve still sums to 1, preserving the total probability. What changes is not the amount of belief, but how that belief is distributed.

In the decades following Bayes’ quiet mathematical insight, it was Pierre-Simon Laplace who expanded this modest idea into a grand framework for reasoning under uncertainty. Between 1774 and 1812, Laplace independently developed and generalized what we now call Bayesian probability.

Laplace introduced the formal concepts of prior (here P(X)), and posterior (probability, giving structure to the idea of updating beliefs in light of new data. He was also the first to express Bayes’ Theorem in its modern algebraic form and apply it rigorously to real-world problems, particularly in astronomy and statistical estimation.

Inverting Probabilities, Updating Beliefs

Bayes’ rule has become the fundamental mathematics of probabilities and an indispensable part of risk modeling. It helps us update what we believe based on what we observe.

For example, if you’re vaccinated and want to know the chance, you’ll still get a disease, Bayes’ rule lets you calculate that by inverting data of how many vaccinated people got sick.

This same approach is central to Bayesian statistics. In traditional (frequentist) statistics, we usually start with a model and ask: If this model is true, how likely is it that we would observe this data? This is useful, but it doesn’t directly tell us how confident we should be in the model itself.

Bayesian statistics inverts this proposition. Using Bayes’ rule, we ask: Given the data we've observed, how likely is it that this model is true? Instead of just asking how likely the data are under a given model? Bayes rules allow us to answer given the data, how likely is the model?

This shift is powerful. It means we can compare different models or hypotheses directly, update our confidence in them as new data arrives. It turns statistics into an evolving dialogue between belief and evidence. We will delve deep into Bayesian statistics in the second part of the course.

From Reverend Bayes to Bayesian Networks

Once we understand the relationship between X and Y, the natural next step can be to follow the chain further: what about a third variable, or a fourth?

If the table and ball experiment taught us how to infer X from Y, we can imagine continuing the sequence. Suppose we now place a wooden bar at position Y, and drop a second ball that lands somewhere position Z < Y. This gives us a new question: Given Z, what can we say about Y? And in turn, what does that tell us about X?

This idea of connected, conditional relationships, where knowledge flows from one variable to another, laid the foundation for connecting large networks of variables in what is known now as Bayesian Networks.

Bayesian Networks

While several scholars had continued exploring this idea, including the notable Professor Andrey Markov in the early 20th century, the formal foundation of Bayesian Networks was proposed more recently, in the 1980s, by Judea Pearl.

Pearl transformed Bayes’ rule from a two-variable equation into a framework for reasoning across entire webs of interconnected uncertainty. In his book, Probabilistic Reasoning in Intelligent Systems (1988), Pearl introduced Bayesian Networks as graphical models that use directed acyclic graphs (DAGs), as shown in Figure 4, to represent dependencies among variables (5).

In his recent research biography, The Book of Why, Pearl suggests that he drew inspiration for message passing through layers and nodes from early advances in neural networks (3).

Pearl, along with other key scholars, played a central role in advancing inference in Bayesian Networks through the development of belief propagation, also known as message passing. This method made it possible to perform efficient probabilistic inference, even in complex systems with many interdependent variables.

Building on this foundation, Philip Dawid (1992) formalized the clique tree representation, which provided a structured way to organize computations across interconnected variables(6). Ross D. Shachter (1986) introduced the tree decomposition method for exact inference, enabling networks to be restructured into simpler forms suitable for computation (7). In a key advance, Steffen Lauritzen and David Spiegelhalter (1988) applied the junction tree framework to Bayesian inference, allowing for local computations instead of costly global ones(8). Finally, Finn Jensen (1994) contributed by developing practical implementations of the algorithm, helping bring these theoretical advances into widespread use (9).

Wider Adoption and Impact

By the 1990s, Bayesian Networks had found their place in AI, expert systems, and decision-support tools.

Microsoft, for instance, used them in its Office troubleshooting assistant. The networks enabled machines to reason in uncertain environments, just as humans do—by updating beliefs, weighing alternatives, and inferring hidden causes from visible effects.

In a 1990s Los Angeles Times article, Bill Gates, then Microsoft Chairman, stated that Microsoft's competitive advantage, particularly in areas like understanding human speech and building internet services, was its expertise in "Bayesian networks".

From a single hidden ball on a table to vast networks modeling disease, risk, or intent, the journey from Bayes’ rule to Bayesian Networks reveals a deep truth: knowledge doesn't exist in isolation—it spreads through structured, conditional relationships. The real power lies in tracing those threads.

Three Fundamental Patterns of Probabilistic Interdependence

As we expand from two to three variables, a new layer of complexity emerges.

There are only three distinct ways to structure three nodes in a directed acyclic graph (DAG), each with a different pattern of causal influence. These three configurations, shown in Figure 5, form the fundamental building blocks of all Bayesian Networks(1).

The first structure is the sequential chain (Figure 5) where A affects C through mediator B. Therefore, conditioned on B, A and C are independent. For example, remoteness, affects the need for new infrastructure, which in turn affects cost. By knowing the exact need for infrastructures, remoteness becomes independent of cost (through that pathway).

The diverging structure in Figure 5 is the declaration of a common cause. The C and B are probabilistically dependent through A, and independently conditioned on A. For example, remoteness, affects the need for new infrastructure, and through another pathway, it affects the labor shortage. In absence of knowledge about the project’s remoteness, it is possible to “induce” that a project with labor shortage, may as well require more infrastructure spending, as it may be remote. However, knowing the degree of remoteness blocks this dependency by creating a conditional independence. Therefore, labour shortage is independent of infrastructure spending when the remoteness is already known. This form of inductive reasoning allows for modeling the latent risk factor (i.e., risk driver) through their effects on measurable risk indicators.

The converging structure in Figure 5 is the declaration of a common effect. The structure exerts that A, and B are probabilistically independent, but both affect C. However, conditioning on C creates probabilistic dependency between A and B. For example, the project profitability is a common effect of both the payback period, and the rate of return. In the absence of knowledge about profitability, the payback period and the rate of return are independent. However, knowing that the project is not profitable creates a conditional dependence between the two. If the rate of return is high, the low profitability of the project can only be explained away by a long and undesirable payback period. Unlike the previous two structures where conditioning on the cause or the mediator creates probabilistic independence, conditioning here on the effect creates dependence.

Together, these three structures—chains, forks, and colliders—are the DNA of Bayesian reasoning. They show how causality, dependence, and inference can shift dramatically depending on where in the network we look—and what we condition on.

Mathematics of Risk

In many instances, risk can be a qualitative and multi-faceted concept, it may involve social judgment, ethical concerns, and strategic trade-offs.

However, when we need to quantify risk for decision-making and modeling, we can break it down into three core elements: a scenario or event that might occur, the likelihood of that event, and the consequence if it happens.

This decomposition gives us the classic risk triplet:

• Si​: a specific scenario or set of events that leads to exposure or harm

• Pi​: the likelihood or frequency of that scenario

• Ci​: the consequence, such as financial loss or operational impact

To assess whether these consequences matter, we introduce another layer: thresholds, denoted Tj. These are the limits of acceptable impact across different domains, 1 to j, such as finance, safety, environment, or reputation. A risk becomes significant, to various degrees, when the expected consequence Pi​ Ci exceeds a relevant threshold Tj​.

At a preliminary maturity level in risk modeling, we might treat the likelihood Pi​ as a fixed probability, and the consequence, Ci, and the threshold, Tj​, as deterministic values (single numbers).

At higher maturity levels, we recognize that all three elements—likelihood, consequence, and threshold—carry uncertainty. These are not always known with precision; they often depend on expert judgment, assumptions, or incomplete and scarce previous data.

Just like the position of the first and second balls on the pool table, Z, and Y, inform us about the uncertainty of the location of the disappeared wooden bar, X. Bayesian methods allow us to work with uncertainties at both levels.

Bayesian methods provide a mathematical framework that allows for two fundamental reasoning capabilities.

1. To model real world phenomena as part of connected systems with probabilistic interdependencies, and,

2. to represent our current understanding and expert judgement as probability distributions and update them as new evidence and data emerges.

Risk Modeling with Bayesian Networks

The classic bow-tie diagrams are useful conceptualizations of risk.

As shown in Figure 6, the bow-tie diagram illustrates a hazardous event at the center, flanked by causes on the left and consequences on the right. It helps visualize how threats may lead to loss of control, and how recovery measures can mitigate consequences.

In conventional risk and reliability engineering, this conceptual bow-tie is often transposed into a combination of Fault Tree Analysis (FTA) and Event Tree Analysis (ETA), as seen in Figure 7. FTA traces backward from a system failure to its root causes through logical gates, while ETA projects forward from a triggering event to multiple outcomes by mapping safety function successes and failures.

While both are foundational tools—logical, binary, and elegantly structured—they struggle to capture probabilistic dependencies beyond the "AND" and "OR" logic gates. They assume that components fail independently unless explicitly connected by logic, and they lack the flexibility to express more subtle interdependencies across time and space.

But dependencies do arise, and they often do so in ways that challenge such linear thinking.

These can manifest through the Three Fundamental Patterns of Probabilistic Interdependence as mentioned in the previous section. This Probabilistic Interdependencies include common causes (e.g., several factors triggering MUE), common effects (multiple failure paths converging on a single point), or serial cause-effect chains (where a risk event triggers a process that in turn creates new vulnerabilities).

Bayesian Networks allow us to introduce probabilistic dependencies across multiple layers of systems and operations.

This graphical structure allows us to do more than just trace risk pathways, it enables both inference and diagnosis by inversing Bayesian probabilities. If we observe a failure, we can use Bayes’ rule to reason backward and identify the most likely cause. Or, if we observe multiple upstream warnings, we can anticipate the probability of a future breakdown.

Bayesian Networks can turn bow-tie diagrams into living systems—ones that learn, update, and support real-time decision-making in uncertain environments.

Bayesian Networks in Projects and Operations

The large-scale adoption of BN in probabilistic risk modeling is mainly because of BNs capability to incorporate incompatible and disparate sources of information, such as data and expert knowledge, as probabilistic beliefs.

Several noteworthy literature reviews have been published for BN applications in risk analysis (10), human reliability assessment (11) , water resources management (12), and ecological risk assessment (13).

McCabe first proposed to model risk in the construction domain in the late 1990s, and used BNs to create probabilistic performance measurements for construction simulation (14–17). Later, researchers used BNs to model project cost risk (18), schedule risk, or both (19).

Noteworthy applications in risk, reliability, and resilience appeared in the literature starting from early 2000s (20). Researchers used BNs to model regulatory compliance risk of different systems (21). BNs were used to model the concept of resilience by expanding reliability to include vulnerability and maintainability (22). An interesting and transferable use to the project domain was modeling the risk of cascading or progressive failure (domino effects) (23).

Concurrently, BNs were used as an actuarial method to model operational risks in financial institutions (24). BNs were successfully utilized to model operational risk with particular emphasis on its ability to conduct scenario modeling for operational risks (25).

The applications of BNs in operational risk overlapped with inclusion of organizational risk, human risks, and eventually socio-technical risks. Researchers proposed frameworks to establish socio-technical risk indicators for BNs (26). The challenge to quantify socio-technical risks was first to combine the social and technical sources of information.

BNs were successfully proposed for nuclear power generator software systems and were adopted by other researcher to create models for the nuclear industry (27–29)

An interesting application of BNs to model organizational risks was proposed to model the probability of collision in maritime transportation (30). Researchers proposed standardized methodologies for probabilistic risk analysis of socio-technical systems (31). BNs were proposed to model diverse sources of risk for complex construction projects (32).

Bayesian Networks in Digital Twins

Certain types of digital twins, especially process, system, and infrastructure-level twins, rely heavily on simulation to model complex behaviors, interactions, and uncertainties.

However, real-time decision-making poses a challenge when relying on simulation. Simulations are computationally intensive and time-consuming, making them unsuitable for instant responses in live systems. By training BNs on the outputs of offline simulations, digital twins can embed probabilistic reasoning models that offer immediate, data-driven insights and predictions.

Dynamic Bayesian network (DBN) framework were used for aircraft digital twin to help in crack growth monitoring, enabling probabilistic diagnosis and prognosis (33). Other researchers extend this by developing a nonparametric Bayesian network-based digital twin, employing Dirichlet process mixture models (DPMM) and Gaussian particle filters to autonomously update model structure and parameters, thus improving accuracy in the health assessment of electro-optical systems (34).

Together, these studies illustrate the value of Bayesian inference in enabling adaptive, interpretable, and computationally efficient digital twins across safety-critical domains.

Question then Becomes

Bayesian Networks are now ubiquitous, silently powering the intelligent systems that surround us.

In generative AI, they enable models to estimate uncertainty, generate realistic data, and adapt intelligently to new evidence. From speech recognition on your phone to diagnostics in complex machinery, BNs are quietly reasoning beneath the surface, helping systems make sense of uncertainty.

• In complex engineering projects, Bayesian Networks are used to model cascading risks, prioritize mitigation strategies, and inform stakeholder decisions under uncertainty.

• In operations, they provide predictive foresight by linking observable indicators to hidden system states, enhancing reliability and responsiveness.

• In digital twins, Bayesian models serve as dynamic reasoning engines—constantly updating beliefs as new data flows in—transforming static simulations into adaptive decision-support systems.

In this article, we traced the journey from Bayes’ original idea to its modern-day applications. We reviewed key concepts like the pool table thought experiment, patterns of probabilistic interdependence, and how BNs help us move beyond simple logic-based risk tools. We explored their use in three major areas: projects; operations; and digital twins.

Despite all this, Bayesian Networks are not a standard part of risk management practices. Why is that?

Is it because today’s project, and operations start with too qualitative, and too complex types of risks, entangled with social and political judgments, to be modeled?

Or is it because building and maintaining Bayesian models requires a mix of technical skill and domain knowledge that’s hard to find and maintain outside of advanced analytics teams of computer scientists?

What will it take for Bayesian thinking to become a practical, everyday tool in managing project and operational risk, just as it already is in the intelligent systems we rely on every day?

Notes

Join the EPM Network to access insights, influence our research, and connect with a community shaping the industry's future.

Subscribe now

Support us by sharing this article with your friends and colleagues, or over social media.

If you wish to share your opinion, provide insights, correct any details in this article, or if you have any questions, please email editor@epmresearch.com.

Refer to this article using the following citation format:

• Zangeneh, P. (2025), “From Reverend Bayes to Modern Risk Modeling in Projects, Operations, and Digital Twins.” Bayesian Methods for Risk Modeling - Part 1, EPM Research Letters.

This article is a reading material for the “Bayesian Methods for Risk Modeling (BMRM)” and “ENCI 604 - Risk, Uncertainty and Reliability” courses at the University of Calgary, Department of Civil Engineering.

References

1. Zangeneh, P. (2021). Knowledge representation and artificial intelligence for management of socio-technical risks in megaprojects [PhD thesis, University of Toronto (Canada)]. ProQuest Dissertations Publishing.

2. Koller, D., & Friedman, N. (2009). Probabilistic Graphical Models: Principles and Techniques. MIT Press.

3. Pearl, J., & Mackenzie, D. (2018). The Book of Why: The New Science of Cause and Effect. Basic Books.​

4. Fenton, N., & Neil, M. (2018). Risk Assessment and Decision Analysis with Bayesian Networks. CRC Press.​

5. Pearl, J. (2014). Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference. Elsevier.​

6. Dawid, A. P. (1992). Applications of a general propagation algorithm for probabilistic expert systems. Statistics and Computing, 2(1), 25–36.​SpringerLink

7. Shachter, R. D. (1986). Evaluating influence diagrams. Operations Research, 34(6), 871–882.

8. Lauritzen, S. L., & Spiegelhalter, D. J. (1988). Local computations with probabilities on graphical structures and their application to expert systems. Journal of the Royal Statistical Society: Series B (Methodological), 50(2), 157–224. http://www.jstor.org/stable/2345762

9. Jensen, F. V. (1997). Introduction to Bayesian Networks. Springer New York.​

10. Weber, P., Medina-Oliva, G., Simon, C., & Iung, B. (2012). Overview on Bayesian networks applications for dependability, risk analysis and maintenance areas. Engineering Applications of Artificial Intelligence, 25(4), 671–682.​ScienceDirect

11. Mkrtchyan, L., Podofillini, L., & Dang, V. N. (2015). Bayesian belief networks for human reliability analysis: A review of applications and gaps. Reliability Engineering & System Safety, 139, 1–16.​

12. Phan, T. D., Smart, J. C. R., Capon, S. J., Hadwen, W. L., & Sahin, O. (2016). Applications of Bayesian belief networks in water resource management: A systematic review. Environmental Modelling & Software, 85, 98–111.​

13. McDonald, K. S., Ryder, D. S., & Tighe, M. (2015). Developing best-practice Bayesian belief networks in ecological risk assessments for freshwater and estuarine ecosystems: A quantitative review. Journal of Environmental Management, 154, 190–200.​

14. McCabe, B., AbouRizk, S. M., & Goebel, R. (1998). Belief networks for construction performance diagnostics. Journal of Computing in Civil Engineering, 12(2), 93–100.

15. McCabe, B., & AbouRizk, S. M. (2001). Performance measurement indices for simulated construction operations. Canadian Journal of Civil Engineering, 28(3), 383–393.

16. McCabe, B., AbouRizk, S. M., & Goebel, R. (1998). Belief networks for construction performance diagnostics. Journal of Computing in Civil Engineering, 12(2), 93–100.

17. McCabe, B., & Ford, D. (2001). Using belief networks to assess risk. In Proceedings of the 2001 Winter Simulation Conference (Cat. No.01CH37304) (pp. 1541–1546). IEEE.

18. Khodakarami, V., & Abdi, A. (2014). Project cost risk analysis: A Bayesian networks approach for modeling dependencies between cost items. International Journal of Project Management, 32(7), 1233–1245.​

19. Lee, E., Park, Y., & Shin, J. G. (2009). Large engineering project risk management using a Bayesian belief network. Expert Systems with Applications, 36(3), 5880–5887.​

20. Mahadevan, S., Zhang, R., & Smith, N. (2001). Bayesian networks for system reliability reassessment. Structural Safety, 23(3), 231–251.​

21. Joseph, S. A., Adams, B. J., & McCabe, B. (2010). Methodology for Bayesian belief network development to facilitate compliance with water quality regulations. Journal of Infrastructure Systems, 16(1), 58–65.

22. Sarwar, A., Khan, F., Abimbola, M., & James, L. (2018). Resilience analysis of a remote offshore oil and gas facility for a potential hydrocarbon release. Risk Analysis, 38(8), 1601–1617.​

23. Khakzad, N., Khan, F., & Amyotte, P. (2013). Risk-based design of process systems using discrete-time Bayesian networks. Reliability Engineering & System Safety, 109, 5–17.​

24. Tripp, M. H., Bradley, H. L., Devitt, R., Orros, G. C., Overton, G. L., Pryor, L. M., et al. (2004). Quantifying operational risk in general insurance companies: Developed by a GIRO working party. British Actuarial Journal, 10(5), 919–1012.

25. Cowell, R. G., Verrall, R. J., & Yoon, Y. K. (2007). Modeling operational risk with Bayesian networks. Journal of Risk and Insurance, 74(4), 795–827.​

26. Øien, K. (2001). A framework for the establishment of organizational risk indicators. Reliability Engineering & System Safety, 74(2), 147–167.​

27. Galán, S. F., Mosleh, A., & Izquierdo, J. M. (2007). Incorporating organizational factors into probabilistic safety assessment of nuclear power plants through canonical probabilistic models. Reliability Engineering & System Safety, 92(8), 1131–1138.​

28. Gregoriades, A., & Sutcliffe, A. (2008). Workload prediction for improved design and reliability of complex systems. Reliability Engineering & System Safety, 93(4), 530–549.​

29. Papazoglou, I. A., Bellamy, L. J., Hale, A. R., Aneziris, O. N., Ale, B. J. M., Post, J. G., et al. (2003). I-Risk: Development of an integrated technical and management risk methodology for chemical installations. Journal of Loss Prevention in the Process Industries, 16(6), 575–591.​

30. Trucco, P., Cagno, E., Ruggeri, F., & Grande, O. (2008). A Bayesian belief network modelling of organisational factors in risk analysis: A case study in maritime transportation. Reliability Engineering & System Safety, 93(6), 845–856.​

31. Léger, A., Weber, P., Levrat, E., Duval, C., Farret, R., & Iung, B. (2009). Methodological developments for probabilistic risk analyses of socio-technical systems. Proceedings of the Institution of Mechanical Engineers, Part O: Journal of Risk and Reliability, 223(4), 313–332.​

32. Qazi, A., Quigley, J., Dickson, A., & Kirytopoulos, K. (2016). Project Complexity and Risk Management (ProCRiM): Towards modelling project complexity driven risk paths in construction projects. International Journal of Project Management, 34(7), 1183–1198.​

33. Li, C., Mahadevan, S., Ling, Y., Choze, S., & Wang, L. (2017). Dynamic Bayesian network for aircraft wing health monitoring digital twin. AIAA Journal.

34. Yu, J., Song, Y., Tang, D., & Dai, J. (2020). A digital twin approach based on nonparametric Bayesian network for complex system health monitoring. Journal of Manufacturing Systems, 57, 206–217.

Highlights

• The construction and project management industry is rapidly evolving due to transformative technologies such as Gen-AI and Extended Reality.

• A survey of twenty industry experts identified three categories of functions impacted by AI: efficiency functions, workflow functions, and human-centric functions.

• Participants believe Gen-AI shows the highest impact in structured tasks such as scheduling, budgeting, and risk management, while Extended Reality enhances collaboration and training.

• The industry anticipates continued AI and XR integration, automation in construction, and the rise of hybrid AI models in 2025.

• The study aims to expand into an annual survey in collaborations with universities and industry leaders worldwide.

The world is undergoing a period of unprecedented transformation, and the construction and project management industry is no exception.

Rapid advancements in technology, coupled with evolving societal and environmental demands, are reshaping how projects are conceptualized, and delivered. Amidst this flux, understanding the role of transformative technologies in enhancing business processes is critical for construction and project management firms, and professionals alike.

We consulted with twenty industry practitioners and experts from our EPM Network, representing different regions across the globe. This report reflects the collective insights and expertise of professionals deeply engaged in the fields of construction and project management, who are using transformative technologies at the forefront of engineering projects.

The results revealed three distinct categories of construction and project management functions being affected. Interestingly, these categories of functions can be explained by their reliance on human activity into efficiency functions, workflow functions, and human-centric functions.

Our experts suggest a future of faster, more streamlined project delivery and development with AI; however, its impact varies across different functions.

This article explores key insights from our study, covering the following topics:

1. Objectives and Audience

2. Transformative Forces in 2024

3. Survey Design

4. Results

   1. Impacts of Generative AI

   2. Impacts of Extended Reality

   3. What to Expect in 2025

5. Major Projects Trends in Canada

6. Discussion: Perception and Reality

7. Conclusion and This Year’s Plan

At the EPM Research Group, we aim to combine rigorous academic research with practical industry insights. We are part of the University of Calgary’s Civil Engineering Department and backed by the Engineering Project Management Endowment.

Objectives and Audience

This study was commissioned as a proof of concept to gauge interest, assess feasibility, and address validity concerns for an annual industry survey program on the impact of transformative technologies on construction and project management.

We asked experts about the impacts of transformative technologies in construction and project management and their opinions on questionnaire design.

Our vision is to create an annual survey to capture the pulse of the industry and its technological transformations year after year. We hope to co-create a resource that informs, inspires, and empowers stakeholders across the industry. The objectives of this survey program are:

• Assessing the current state of industry adoption of transformative technologies.

• Understanding technology adoption across various sub-sectors of the construction industry and across different countries and regions.

• Exploring professional perspectives on the potential, limitations, and challenges of these technologies.

• Examining the industry’s outlook on disruptive transformations.

• Guiding academic research by highlighting industry’s emerging needs.

We defined the sample population as industry practitioners (both governmental agencies and private industry) in construction and project management from around the world who are at the forefront of leveraging transformative technologies in their projects.

The study benefits a diverse audience in construction and project management, from engineers and project managers to business leaders, academics, policymakers, and young engineers looking to navigate their careers.

This range of audiences and stakeholders reflects the interconnected nature of our industry.

Transformative Forces in 2024

We focused on two transformative technologies: Generative AI and Extended Reality.

These technologies were selected for their significant emergence in 2024 and their potential to reshape the construction and project management landscape.

Generative AI

Generative Artificial Intelligence (Gen-AI)—encompassing technologies such as Large Language Models (LLMs) and Large Image-Generative Models (LIGMs)—has emerged as a transformative force, redefining industries at an unprecedented pace.

The rapid ascent of Generative AI is affecting both academic research and technology development across many industrial sectors.

Gen-AI distinguishes itself by its ability to create new content—text, images, and code—that mirrors its training data. Unlike traditional AI, which primarily analyzes or classifies, Gen-AI extends beyond existing data by generating novel outputs.

The adoption of Gen-AI in construction and project management remains in its early stages, often limited to proof-of-concept applications [1]. Several challenges hinder widespread adoption, including validating AI-generated outputs, fine-tuning models for domain-specific applications, mitigating biases, ensuring data security, and addressing ethical implications.

In 2024, OpenAI introduced GPT-4o, a multimodal model capable of processing and generating text, images, and audio in real-time, setting new benchmarks in voice recognition and translation. The year also marked a dual surge in the rise of open-source models and the adoption of smaller, more efficient AI systems [2].

Open-source releases, such as Meta’s LLaMA 3 and Databricks’ DBRX, aimed to democratize AI and challenge the dominance of proprietary giants like OpenAI [3,4]. Alibaba joined this movement by releasing over 100 open-source models, including text-to-video technology, expanding AI’s applications in automotive, gaming, and scientific research [5].

In parallel, companies such as DeepSeek are reducing training costs through innovative methods, including the Mixture-of-Experts (MoE) approach, which routes tasks to specialized expert models for greater efficiency. This technique enables AI models to be trained at a fraction of the cost compared to traditional methods [6].

Simultaneously, the industry is embracing compact, task-specific models designed for efficiency and affordability, prioritizing practical deployment and accessibility across a range of applications [7].

Extended Reality

Extended Reality (XR)—encompassing Virtual Reality (VR), Augmented Reality (AR), and Mixed Reality (MR)—was another transforming force in 2024, aiming to change the way we work, learn, and connect by merging physical and digital spaces.

Despite hardware limitations, XR remains at the forefront of technological development due to its potential to revolutionize collaboration, training, and problem-solving. By enabling shared virtual spaces, XR allows teams to brainstorm, simulate, and interact with digital objects in real time, fostering seamless collaboration across virtual landscapes. XR provides hyper-realistic simulation environment for skill development and risk-free scenario testing [8].

In construction and project management, XR can transform training, remote operations, and stakeholder collaboration [9]. Immersive simulations allow workers to practice operating machinery, navigate complex sites, and respond to emergencies within risk-free, hyper-realistic environments [10,11]. Virtual offices enable stakeholders to explore digital twins of projects, refine designs, and collaboratively resolve physical and logistical challenges in real time [12].

The year 2024 marked a significant leap in XR advancements. Apple introduced the Vision Pro [13], a mixed-reality headset that seamlessly blends digital and physical worlds, offering professionals a powerful tool for reimagining project visualization and collaboration.

Not to be outdone, Meta unveiled the Orion glasses [14], lightweight AR eyewear that overlays virtual objects onto the real environment, enhancing on-site decision-making and stakeholder engagement. Google entered the arena with Android XR [15], an operating system tailored for XR devices, paving the way for a new wave of applications that could revolutionize how we approach construction projects.

Survey Design

The survey structure for this study consisted of three main components: Gen-AI Impact, Extended Reality Impact, and Industry Outlook.

Before answering the questions, participants were provided with descriptions of the strengths and weaknesses of Gen-AI and Extended Reality based on a systematic literature review we conducted.

1- Generative AI Impact

Participants evaluated the potential for productivity gains that Generative AI could bring to 14 functions and processes within construction and project management.

Each area was rated on a scale from 1 to 9, with 9 indicating the highest potential for productivity gains. The 14 functions and processes within construction and project management included are as follows:

1. Project Planning and Scheduling: Streamlining timelines, resource allocation, and schedule resilience.

2. Budgeting and Cost Estimation: Automating cost forecasts and refining financial accuracy.

3. Risk Management: Enhancing risk identification and mitigation strategies using AI-driven insights.

4. Project Controls: Improving monitoring, execution, and corrective actions.

5. Procurement and Supply Chains: Optimizing vendor selection and contract management processes.

6. Contract Management and Tendering: Streamlining bid evaluations and contract negotiations.

7. Layout Planning and Logistics: Advancing site layout designs and resource movement efficiency.

8. Quality Control and Compliance: Automating quality assurance processes and ensuring regulatory adherence.

9. Stakeholder Engagement: Personalizing communication strategies and improving stakeholder alignment.

10. Lessons Learned and Continuous Improvement: Enabling systematic project close out reports, reviews, and knowledge-sharing.

11. Human Resources and Staffing: Enhancing workforce planning and skills development.

12. Occupational Health and Safety: Leveraging AI to monitor and improve safety measures.

13. Trend and Change Management: Managing project change orders and their impact dynamically.

14. Conflict Resolution: Facilitating proactive identification and resolution of disputes.

These fourteen construction and project management functions were described in greater detail in the survey compared to this report. The descriptions helped to establish a common baseline of understanding.

2- Extended Reality Impact

The survey included targeted questions about the familiarity of participants with Extended Reality (XR) technologies. This set of questions aimed to understand the current level of adoption and firsthand experience with XR technologies in the industry.

Next, participants were asked which of the previous functions in construction and project management would most benefit from the convergence of Generative AI and XR. The question further encouraged respondents to reflect on the practical applications of combining AI-driven insights with immersive XR tools.

3- Industry Outlook

Lastly, the participants were asked to reflect on their outlook for the coming year. This question was further discussed with some of participants over conference calls to grasp their point of view on the technological trends.

Results

In total, twenty industry experts participated, offering perspectives on the areas where Generative AI might exert the greatest influence. The respondents were asked to rank fifteen critical domains in construction and project management on a scale from 1 to 9, with 9 indicating the highest potential and 1 reflecting minimal impact.

Impacts of Generative AI

The results, visualized in the spider plot of Figure 1, provide a clear snapshot of how industry experts perceive the varying impacts of Generative AI in productivity gains across different construction and project management functions.

We refrained from employing more time-intensive pairwise comparisons in favor of simple Likert scales to keep the survey efficient for participants. Despite this simplification, the results revealed distinct patterns in the data, showing clear variations in the perceived impact of generative AI across functions.

Clustering of Generative AI Impact Areas

We identified three distinct groupings based on respondents' perceptions of the relative potential for productivity gains.

These groupings allowed us to classify construction and project management functions into three categories:

• Efficiency Functions, which saw the highest perceived impact (e.g., planning, scheduling, and cost estimation);

• Workflow Functions, reflecting moderate potential (e.g., logistics, procurement, and compliance); and,

• Human-Centric Functions, where the impact was perceived to be lower due to the reliance on interpersonal skills and decision-making (e.g., stakeholder engagement and conflict resolution).

Figure 2 depicts this categorization and the areas where generative AI can complement and enhance current practices.

1. Efficiency Functions: Generative AI demonstrated the highest potential in domains relying on structured workflows, repetitive updating, and large datasets. Participants saw immediate value in leveraging AI to increase productivity on these activities.

   • Project Planning and Scheduling: Automating schedule templates, identifying potential bottlenecks, and analyzing historical data.

   • Budgeting and Cost Estimation: Accelerating conceptual cost estimating, cost breakdown generation, and anomaly detection.

   • Risk Management and Project Controls: Generating predictive insights, improving risk analysis, and creating automated dashboards.

   • Contract Management and Tendering: Automating contract drafting, vendor analysis, and tender evaluations.

   • Lessons Learned: Compiling and analyzing project histories to identify best practices and inform future strategies.

2. Workflow Functions: The synergy between AI tools and human judgment was emphasized due to the often time intensive workflow and case specific context of these functions. These domains benefited from AI's ability to handle routine, data-intensive tasks while relying on humans to navigate the process:

   • Procurement and Supply Chains: Automating RFPs and supplier evaluations while preserving human oversight for negotiations.

   • Layout Planning and Logistics: Proposing site layouts and logistics scenarios, adaptable by human planners.

   • Quality Control and Compliance: Standardizing checklists and identifying gaps, with human-led physical inspections.

   • Human Resources and Staffing: Streamlining applicant processing while retaining human interactions for cultural fit assessments.

   • Project Trends and Change Management: Identifying trends and suggesting changes while requiring expert integration.

   • Occupational Health and Safety: Drafting safety policies with humans ensuring environment-specific adaptation.

3. Human-Centric Functions: These domains—heavily influenced by interpersonal skills and real-time decision-making—showed moderate or lower potential for AI adoption. While AI can assist with tools like sentiment analysis and automated reports, the human element remains crucial:

   • Stakeholder Engagement: Supporting sentiment summaries but maintaining trust-building as a human-led activity.

   • Conflict Resolution: Analyzing conflict data to suggest strategies, yet relying on empathy and nuanced judgment for resolutions.

While this preliminary survey respondent pool was limited to only twenty experts, it highlighted an interesting delineation between tasks where AI can lead, collaborate, or support.

Impacts of Extended Reality

After exploring the perceived impact of Generative AI, we assessed participants’ familiarity with Virtual Reality (VR) and Augmented Reality (AR) technologies.

About 40% of the participants reported having used these tools, highlighting relatively limited adoption of XR in construction and project management.

Despite limited exposure to XR technologies, respondents demonstrated significant optimism about the potential for combining XR with Generative AI to address critical industry challenges. This convergence offers unique opportunities in areas such as project controls, 4D/5D scheduling, risk management, layout planning, and procurement. Figure 3 highlights the perception of participants as to which function stands to gain the most from integration of XR and Generative AI.

A common theme in our conversations was around how by harnessing the immersive capabilities of XR, professionals can simulate AI-generated scenarios, enabling them to visually explore and interact with project data in real-time.

What to Expect in 2025

Lastly, participants were asked to share their outlook for the coming year, with follow-up discussions conducted through one-on-one calls to better understand their perspectives on technological trends.

At the Engineering Project Management (EPM) research group, our research is focused on three of these transformative themes shaping the future of construction and project management, marked below with an asterisk: the integration of AI and XR technologies, Hybrid ML-LLM AI models, and the adoption of Small Language Models (SLMs) with edge computing.

1- AI and XR Integration*

The integration of AI and XR has emerged as a key trend, identified by survey respondents and widely recognized across the industry.

Meta has addressed the demand for user-friendly devices with the Ray-Ban Meta Smart Glasses, a lightweight option featuring a camera and Generative AI capabilities, though it lacks full XR functionality [16]. This trade-off prioritizes wearability and everyday use. Looking ahead, Meta is developing Orion AR glasses, which promise advanced AR and AI integration in a sleek form [14]. Expected by late 2027, these glasses reflect Meta's focus on overcoming current limitations in wearable technology [4].

2- AI in Automation and Robotics

By 2025, AI-driven automation and robotics are set to transform the construction industry, streamlining processes and enhancing safety. Advanced robotics will take on complex tasks with unmatched precision and speed, reducing physical strain on workers and mitigating risks in hazardous environments [17].

Autonomous drones equipped with AI will revolutionize site management by conducting inspections, monitoring progress, and identifying potential issues like structural weaknesses or material shortages in real time. These insights will enable project managers to optimize resources, maintain schedules, and make data-driven decisions [18].

Tesla is advancing AI robotics with innovations like 'Optimus,' which could be adapted for construction tasks. Such robots, capable of performing diverse manual activities, have the potential to reduce labor-intensive work and further expand the scope of automated construction technologies [19].

3- Hybrid ML-LLM AI Models*

LLMs excel in capturing linguistic patterns, generating coherent text, and generalizing across tasks. Their ability to perform hierarchical task planning through analogical reasoning has become central to LLM research. However, traditional AI and machine learning models surpass LLMs in tasks requiring accuracy and prediction. By integrating LLMs' task decomposition with ML data-driven forecasting, a hybrid ML-LLM AI approach offers transformative solutions in construction and project management.

ML can analyze historical project data to predict risks, optimize processes, and identify patterns that lead to inefficiencies, enabling proactive management [20]. LLMs communicate these results with diverse teams, automate documentation, and generate compliance reports and proposals, while reducing administrative effort [21,22].

4- Open Source / Small Language Models (SLM) and Edge Computing*

Open Source and Small Language Models (SLMs) represent a transformative opportunity for construction and project management, offering cost-effective, targeted solutions to manage complex data and communication flows.

Models like GPT-4o-mini and Gemini-flash provide a streamlined alternative to Large Language Models (LLMs), enabling organizations to deploy powerful natural language processing applications with significantly reduced investment [7].

Beyond cost efficiency, the value of SLMs lies in their adaptability; they can be refined and deployed on local servers disconnected from the broader internet which is an immense advantage when dealing with proprietary information and sensitive project data.

SLMs are also well-suited for integration with XR (Extended Reality) technologies, which often operate on portable devices with limited processing power. This synergy facilitates real-time, context-sensitive support through augmented reality overlays and virtual environments, enhancing operational efficiency as well as improving training and safety protocols [23].

5- Generative AI Design and Architecture

A few experts in our survey population were eagerly monitoring the impact of Generative AI on architectural design in 2025.

The Large Image Generation Models (LIGM) are set to significantly expanding the way architects approach the creation and refinement of building concepts. This cutting-edge technology enables the input of design parameters and constraints into AI systems, which then generate a wide array of innovative design options [24].

6- Avatars Everywhere

The concept of avatars—digital personas that represent users in virtual environments—has evolved significantly, becoming a versatile tool in various industries. In 2025, these digital representations are not only personal avatars in social media or gaming but have expanded into professional fields, offering a unique way to interact within digital spaces. Avatars can be customized to mirror the user’s physical appearance or embody a completely designed persona, equipped with realistic expressions and movements [25].

Major Projects Trends in Canada

Understanding major transformative trends in construction and project management benefits from a parallel awareness of the broader landscape of construction project prevalence.

Figure 4 and 5 show the annual trend in project numbers and capital costs for both the resource project and infrastructure projects in Canada as disclosed by Government of Canada open data initiative [26,27].

In terms of the number of projects in Canada, both resource and infrastructure projects have been on the rise in 2024. However, while capital expenditure for resource projects has shown a steady increase, infrastructure investment has declined from 2023 to 2024.

As shown in Figure 4, the energy sector, which remains a cornerstone, has seen consistent activity in electricity generation, oil and gas, and other projects such as biomass, biofuel, and geothermal production, with 340 projects contributing CA$510 billion by 2024. Mining, has exhibited robust growth, with investments growing from CA$89 billion to CA$117.1 billion in four years, driven by metals and an expanding focus on non-metals and critical resources.

Canada's infrastructure investment trends highlight a strategic shift between volume and value. In 2021, the number of projects (excluding those related to COVID 19 response) peaked at 1,377, supported by a record CA$20.94 billion in eligible costs, reflecting a major post-pandemic recovery push. While project counts dropped in subsequent years—reaching 643 in 2022 and 486 in 2023—investment levels remained significant. By 2024, project counts rose again to 802, supported by CA$5.77 billion in funding.

Discussion: Perception and Reality

The rapid evolution of AI continues to reshape industries.

The recent emergence of powerful open-source models by DeepSeek in early 2025 challenged the consensus on proprietary models, highlighting the volatility of the Gen-AI scene and the infancy of this technological race [28].

AI's trajectory remains one of continuous advancement, with no definitive winner yet, as new models and capabilities emerge at an accelerating pace.

As suggested by a curious paradox known as Jevons Paradox, technological efficiency does not always lead to reduced demand [29]. On the contrary, as seen with the steam engine and many other era-defining inventions, improvements in efficiency often result in increased overall demand and consumption.

In construction and project management, the introduction of Gen-AI does not appear to be replacing workers yet. However it has helped in eliminating bottlenecks and increasing the quality and speed of project delivery.

The limited consultations we had with industry experts through this study suggest that instead of seeing lower employment, firms are facing a greater need for skilled professionals. The overall rise in the number of major projects announced in Canada in 2024 may indicate a trend that aligns with efficiency gains.

The reliability of Generative AI models remains a key consideration. Many of the most sensitive functions in construction and project management carry significant liabilities and require human oversight. Gen-AI models can drastically reduce the cost of first drafts, but final work products require accountability that must remain with experienced professionals.

For XR technologies, hardware limitations, including integration challenges and high costs, remain key bottlenecks. Addressing these issues is essential for realizing XR's full potential and achieving seamless adoption across business processes.

Looking ahead, the challenge is not whether AI will reshape the industry—it already is—but how organizations will adapt to these changes.

Conclusion and Next Year’s Plan

Looking ahead to 2025, transformative advancements in AI and XR are expected to drive significant changes in the construction and project management industry, reshaping workflows and enhancing the integration of digital and physical environments.

Our goal for 2025 is to expand the scope of this study, both in reach and depth, into an industrywide survey of the construction and project management industry. We have established contacts with several universities around the world to partner on this survey.

We aim to engage a wider audience across diverse geographies and disciplines to ensure the findings accurately reflect the industry's dynamic and interconnected nature.

Building on the insights gained from this study, we are refining our approach and designing a more precise and expressive questionnaire to capture nuanced perspectives on emerging trends within the construction and project management sector.

If you are interested in supporting this effort and sharing your insights on this year’s or next year’s survey, we would greatly value your contribution. Your perspectives are important for informing the design of next year’s survey. We invite you to connect with us.

Notes

Join the EPM Network to access insights, influence our research, and connect with a community shaping the industry's future.

Subscribe now

Support us by sharing this article with your friends and colleagues, or over social media.

If you wish to share your opinion, provide insights, correct any details in this article, or if you have any questions, please email editor@epmresearch.com.

Refer to this article using the following citation format:

• Zangeneh, P. and Ghorab, Kh. (2025), “Transformative Technologies in Construction and Project Management: 2025 Outlook and Survey Results”, EPM Research Letters.

References

1. Taiwo, R., et al. (2024). Generative AI in the construction industry: A state-of-the-art analysis. arXiv. https://doi.org/10.48550/arXiv.2402.09939

2. Ueno, L., & Lynn, T. (n.d.). GPT-4o: The comprehensive guide and explanation. Retrieved from https://blog.roboflow.com/gpt-4o-vision-use-cases/

3. Grattafiori, A., et al. (2024, November 23). The Llama 3 herd of models. arXiv. https://doi.org/10.48550/arXiv.2407.21783

4. Databricks. (n.d.). Introducing DBRX: A new state-of-the-art open LLM.

5. Wang, P., et al. (2024, October 3). Qwen2-VL: Enhancing vision-language model’s perception of the world at any resolution. arXiv. https://doi.org/10.48550/arXiv.2409.12191

6. Dai, D., et al. (2024, January 11). DeepSeek MoE: Towards ultimate expert specialization in mixture-of-experts language models. arXiv. https://doi.org/10.48550/arXiv.2401.06066

7. Nguyen, C. V., et al. (2024, October 25). A survey of small language models. arXiv. https://doi.org/10.48550/arXiv.2410.20011

8. Adami, P., et al. (2023). Participants matter: Effectiveness of VR-based training on the knowledge, trust in the robot, and self-efficacy of construction workers and university students. Advanced Engineering Informatics, 55, 101837. https://doi.org/10.1016/j.aei.2022.101837

9. Xu, Z., Feng, Z., Babaeian Jelodar, M., & Guo, B. H. W. (2024). Augmented reality applications in construction productivity: A systematic literature review. Advanced Engineering Informatics, 62, 102798. https://doi.org/10.1016/j.aei.2024.102798

10. Pooladvand, S., Taghaddos, H., Eslami, A., Nekouvaght Tak, A., & Hermann, U. (2021). Evaluating mobile crane lift operations using an interactive virtual reality system. Journal of Construction Engineering and Management, 147(11), 04021154. https://doi.org/10.1061/(ASCE)CO.1943-7862.0002177

11. Chen, T., Yabuki, N., & Fukuda, T. (2024). Mixed reality-based active hazard prevention system for heavy machinery operators. Automation in Construction, 159, 105287. https://doi.org/10.1016/j.autcon.2024.105287

12. Sai, S., Sharma, P., Gaur, A., & Chamola, V. (2024). Pivotal role of digital twins in the metaverse: A review. Digital Communications and Networks. Retrieved February 2, 2025, from https://www.sciencedirect.com/science/article/pii/S2352864824001706

13. Apple. (n.d.). Apple Vision Pro. Retrieved January 5, 2025, from https://www.apple.com/apple-vision-pro/

14. Meta. (n.d.). Introducing Orion, our first true augmented reality glasses. Retrieved January 5, 2025, from https://about.fb.com/news/2024/09/introducing-orion-our-first-true-augmented-reality-glasses/

15. Android. (n.d.). Learn more about Android XR. Retrieved February 2, 2025, from https://www.android.com/xr/

16. Meta. (n.d.). Introducing the new Ray-Ban | Meta smart glasses. Retrieved January 5, 2025, from https://about.fb.com/news/2023/09/new-ray-ban-meta-smart-glasses/

17. Wanna, S., Parra, F., Valner, R., Kruusamäe, K., & Pryor, M. (2024). Unlocking underrepresented use-cases for large language model-driven human-robot task planning. Advanced Robotics, 38(18), 1335–1348. https://doi.org/10.1080/01691864.2024.2366974

18. Waqar, A., et al. (2023). Modeling relation among implementing AI-based drones and sustainable construction project success. Frontiers in Built Environment, 9, 1208807. https://doi.org/10.3389/fbuil.2023.1208807

19. Malik, A. A., Masood, T., & Brem, A. (2023, April 11). Intelligent humanoids in manufacturing to address worker shortage and skill gaps: Case of Tesla Optimus. arXiv. https://doi.org/10.48550/arXiv.2304.04949

20. Erfani, A., & Cui, Q. (2022). Predictive risk modeling for major transportation projects using historical data. Automation in Construction, 139, 104301. https://doi.org/10.1016/j.autcon.2022.104301

21. Ghimire, P., Kim, K., & Acharya, M. (2024). Opportunities and challenges of generative AI in the construction industry: Focusing on adoption of text-based models. Buildings, 14(1), 220. https://doi.org/10.3390/buildings14010220

22. Ghimire, P., Kim, K., & Acharya, M. (2023). Generative AI in the construction industry: Opportunities & challenges. arXiv. https://doi.org/10.48550/arXiv.2310.04427

23. Xu, F., Zhou, T., Nguyen, T., Bao, H., Lin, C., & Du, J. (2025). Integrating augmented reality and LLM for enhanced cognitive support in critical audio communications. International Journal of Human-Computer Studies, 194, 103402. https://doi.org/10.1016/j.ijhcs.2024.103402

24. Li, C., Zhang, T., Du, X., Zhang, Y., & Xie, H. (2024). Generative AI models for different steps in architectural design: A literature review. Frontiers of Architectural Research. Retrieved February 3, 2025, from https://www.sciencedirect.com/science/article/pii/S209526352400147X

25. Ante, L., Fiedler, I., & Steinmetz, F. (2023). Avatars: Shaping digital identity in the metaverse.

26. Natural Resources Canada. (n.d.). Natural resources: Major projects planned or under construction 2024 to 2034. Retrieved February 3, 2025, from https://natural-resources.canada.ca/science-and-data/data-and-analysis/natural-resources-major-projects-planned-or-under-construction-2024-2034/26583

27. Infrastructure Canada. (n.d.). Housing, infrastructure and communities Canada - Housing and infrastructure project map. Retrieved February 3, 2025, from https://housing-infrastructure.canada.ca/gmap-gcarte/index-eng.html

28. DeepSeek-AI. (2025, January 22). DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning. arXiv. https://doi.org/10.48550/arXiv.2501.12948

29. York, R., & McGee, J. A. (2015). Understanding the Jevons paradox. Environmental Sociology, 2(1), 77–87. https://doi.org/10.1080/23251042.2015.1106060

NextStar is the largest investment in Canadian automotive history. With a projected annual capacity of 49.5 GWh, enough to power 450,000 EVs. This facility represents a major shift in the Canada’s approach to sustainable transportation and it marks an important milestone in the transition toward cleaner mobility (1).

The global move toward electric vehicles is reshaping industries and economies, and at the heart of this transition lies battery technology. Windsor, Canada’s “automotive capital,” has faced its share of challenges over the years, from shifts in manufacturing to the impacts of economic downturns. Yet, through resilience and adaptability, the city is positioning itself at the forefront of the EV revolution (2).

The NextStar project is expected to create up to 2,500 jobs, offering new opportunities in advanced manufacturing and reinforcing Windsor’s role in the evolving automotive landscape. More than just an economic boost, this initiative reflects a broader commitment to innovation and sustainability in Canadian industry.

This article explores the several dimensions of the NextStar project:

1. Project Highlights, Timeline, and Team

2. Financing Structure

3. Risks and Complexities

4. Lithium-Ion Pouch Type Batteries

5. Project’s Social Impacts

6. Sustainable Development

7. Windsor’s Automotive Legacy

Project Highlights, Timeline, and Team

The NextStar Energy facility spans up to 4.5 million square feet on a 220-acre site. At its core, the facility consists of two primary structures:

• Cell Building (2.7 million SF): Dedicated to manufacturing lithium-ion battery cells, ensuring high-capacity energy storage for EVs.

• Module Building (445,000 SF): Responsible for assembling battery cells into modules, completing the final product before distribution.

The supporting infrastructure includes a master substation, an electrolyte supply system, safety and waste management units, and solvent recovery systems—all designed to enhance efficiency, safety, and sustainability in battery production.

The main production facilities feature high-precision cleanroom and dry room environments to maintain strict humidity and temperature controls, alongside integrated module assembly lines that ensure efficient production and quality assurance.

The project was reported to include at least 38,000 tons of structural steel supplied, along with 3,100 tons of steel joists and 420,000 square feet of decking. The mechanical, HVAC, and utility systems are equally extensive, including 5 million pounds of sheet metal for ventilation and air control, more than 75 miles of hydronic and plumbing pipes, and over 900 HVAC units to maintain precise climate conditions essential for battery production.

Project Timeline

The project was reported to have reached approximately 30% completion by December 2023, with the Module Building fully enclosed and structural steel installation completed. At that time, equipment installation had begun, while the Cell Building was nearing completion of its structural framework, with around 40% of its exterior enclosed (4).

The first phase of operations, focusing on battery module production, was reported to be on track for launch in the first half of 2024. Cell manufacturing, a critical component of the facility’s output, was projected to begin in 2025, with full production expected by 2026 (5).

NextStar reported to have officially commenced battery module production in October 2024, marking a milestone in the project’s transition from construction to operation. The project has also been making strides in workforce development, with over 450 of the projected 2,500 jobs reportedly filled (6).

Project Team

The execution relied on a diverse network of stakeholders, including industry leaders, contractors, engineers, and suppliers. At its core, NextStar Energy, a joint venture between Stellantis and LG Energy Solution drives the plant’s development and oversees its construction and operations.

Sponsors:

• Stellantis: Global automaker and co-founder of NextStar Energy, responsible for overseeing battery production and integration into its EV supply chain(7).

• LG Energy Solution: Leading battery manufacturer, co-founder of NextStar Energy, bringing expertise in lithium-ion battery production (8).

• NextStar Energy: The joint venture entity responsible for the construction, management, and operation of the Windsor EV battery plant (3).

Contractors, Engineers, and Suppliers:

• Alberici-Barton Malow (A-BM): A joint venture comprised of Alberici and Barton Malow is the Design-builder responsible for the construction of the 4-million-square-foot facility, including preconstruction, steel fabrication, and concrete work (9).

• Kasian Architecture Interior Design and Planning Ltd.: Architect of record for the Module Building, responsible for architectural design in collaboration with Gresham Smith, Stantec (Formerly Morrison Hershfield), and A-BM (10).

• Gresham Smith: Architecture firm collaborating on the Module Building design for NextStar Energy. Structural Engineers (11).

• Stantec (Formerly Morrison Hershfield): Engineering firm involved in various aspects of the facility’s development. (12).

• SMRT Architects and Engineers: Providing cleanroom and dry room envelope design and engineering services in collaboration with Hodess Cleanroom Construction (13).

• Hodess Cleanroom Construction: Specializing in cleanroom construction for battery manufacturing processes (13,14).

• Black & McDonald: Responsible for mechanical and HVAC systems, including installing 5 million pounds of sheet metal, 75+ miles of hydronic and plumbing pipe, and 900 HVAC units (15).

• Walters Group: Tasked with installing 3,100 tons of steel, joists, and 420,000 square feet of decking within a four-month timeframe (16).

• Hillsdale Fabricators: A division of Alberici, supplying 38,000 tons of structural steel for the facility (9,17).

Financing Structure

NextStar Energy reflects a strategic collaboration between industry and government, aligning private investment with public incentives.

At the core of the funding structure is a public-private partnership, where government contributions are tied to production milestones. Stellantis is investing CA$3.4 billion in plant construction, while the federal and Ontario governments have each pledged CA$500 million toward development costs. However, much of the government subsidies are conditional on the facility meeting specific production targets, ensuring that public funding is linked to tangible economic and environmental returns (18).

Based on LG Energy Solution's financial statements as of June 30, 2024, the company has committed to investing US$1.464 billion in the joint venture with Stellantis and has already contributed US$765 million. The agreement includes a clause that allows either party to purchase the other's shares in the event of a default, ensuring mutual commitment to the project's success (8).

Risks and Complexities

Project Completion and Ramp Up

Executing a project of this scale comes with a unique set of challenges, requiring advanced design strategies, global coordination, and sophisticated logistics to ensure smooth construction and production.

Innovative Design and Construction: To meet the project's ambitious timeline, prefabrication and digital project management tools have played a crucial role in optimizing efficiency. One of the project stakeholders highlighted the use of Navis, Revit, BIM 360, Procore, Blue Beam, and P6 in their workflow (15). By prioritizing off-site prefabrication, the project has been able to reduce onsite labor demands and accelerate field installation, a crucial factor in meeting the tight construction schedule.

Global Coordination: Given the number of architecture firms, consultants, and stakeholders involved—many spreads across multiple countries—seamless communication and precise coordination are critical to maintaining progress. As the facility ramps up operations, NextStar Energy is collaborating with LG’s facilities in Poland to supply interim battery cells, ensuring production continuity while local manufacturing scales up.

Sophisticated Logistics and Manufacturing Risks: The facility is designed with advanced logistical systems to optimize the movement of raw materials and finished battery cells. However, manufacturing EV batteries at scale presents stringent quality control challenges.

Supply Chain Vulnerabilities

The EV battery supply chain is highly complex and geographically concentrated, making it susceptible to disruptions.

A significant portion of battery-grade materials and components are processed in a limited number of countries, notably China, which controls a substantial share of the global battery supply chain.

In early 2025, the Trump administration proposed a 25% tariff on imports from Canada and Mexico, including automotive parts and metals essential for EV manufacturing (19). This policy aimed to encourage domestic production but introduced challenges for the North American EV industry, which relies on a seamless cross-border supply chain. The tariffs will lead to increased production costs, potential delays, and uncertainty within the industry.

Automakers and suppliers are now reassessing their supply chain strategies, considering relocating production to the U.S. or other regions to mitigate the impact of these tariffs. The lack of regulatory stability is a major risk to continued economic success and thriving of EV battery projects in north America.

Economic and Financial Challenges

The financial viability of EV battery plants is influenced by fluctuating demand for electric vehicles, high production costs, and intense global competition.

The collapse of major European EV battery producer Northvolt in November 2024 underscores the financial risks inherent in the industry, where even well-funded companies can face insolvency due to production challenges and market dynamics (20).

Technology Obsolescence

The rapid evolution of battery technology presents a fundamental challenge for EV battery plants, where large-scale investments in current production methods may risk becoming obsolete as new, more efficient, and cost-effective solutions emerge.

Innovations such as solid-state batteries and lithium-sulfur chemistries are being actively explored to enhance energy density and reduce reliance on scarce raw materials. Companies like Lyten are making significant strides in lithium-sulfur battery production, aiming to disrupt traditional supply chains and improve performance. However, while these next-generation technologies hold promise, they also face hurdles, including technical barriers, high development costs, and the need for new manufacturing processes—all of which could affect the long-term viability of existing battery production facilities (21).

CEO Danies Lee recently emphasized the NextStar Energy plant’s flexibility within lithium-ion technology, highlighting LG Energy Solution’s track record in navigating advancements in battery chemistry over the past two decades. “We’re pretty flexible under the umbrella of lithium-ion battery technology,” Lee noted. “When it comes to the lithium process and chemistry, we can respond to change” (5).

The Windsor plant has been designed with adaptability in mind, allowing for simultaneous production of different battery types while ensuring that its cells remain versatile across multiple vehicle platforms. This strategic approach reflects an industry-wide recognition that resilience and innovation must go hand in hand—balancing the demands of today’s market with the breakthroughs of tomorrow (5).

Lithium-Ion Pouch Type Batteries

NextStar Energy’s decision to manufacture pouch-type lithium-ion batteries marks a shift in the electric vehicle (EV) battery sector, leveraging their unique advantages in space efficiency, energy density, and safety (22).

Unlike traditional cylindrical or prismatic lithium-ion batteries, pouch batteries utilize a flexible aluminum foil casing rather than a rigid metal enclosure, allowing for customizable shapes that optimize space within EV battery packs. This design enhances energy efficiency and reduces weight, a crucial factor for improving EV range.

The manufacturing process, although complex, includes a cutting-edge “degassing” method developed by LG Energy Solution, which improves efficiency and reduces the common issue of electrolyte leakage. In 2022, pouch-type batteries accounted for 26% of the market share in the EV battery industry and are projected to increase to 31% by 2030 due to their high energy density and flexible applications (23).

Unlike lithium-sulfur (Li-S) batteries, which utilize sulfur-based cathodes for higher energy density but suffer from a shorter cycle life, pouch-type batteries rely on conventional lithium-ion chemistry with a liquid electrolyte.

Additionally, all-solid-state batteries, a next-generation alternative, replace the liquid electrolyte with a solid electrolyte, enhancing safety by eliminating leakage risks and improving longevity. While pouch-type batteries still contain liquid electrolytes and require advanced safety measures, their design flexibility, high energy density, and thermal efficiency make them a preferred choice for next-generation EVs, supporting the industry's shift toward lighter, more efficient, and sustainable energy storage solutions (23).

Manufacturing of Lithium-Ion batteries

Lithium-ion batteries generate electricity by moving lithium ions between the cathode and anode during charging and discharging.

The cathode determines capacity and voltage, while the anode, typically graphite, influences lifespan due to structural wear over cycles.

Research into silicon anodes aims to improve capacity and efficiency. The electrolyte, available in liquid, solid, or gel forms, enables ion transport and must maintain high conductivity and stability. A separator, usually made of polyethylene or polypropylene, prevents direct contact between the cathode and anode while allowing ion flow and shutting down at high temperatures. Efforts to develop thinner separators seek to enhance energy density and reduce battery size. Manufacturing process of Lithium-Ion batteries entails four main steps:

1. Electrode Manufacturing: Mixing, coating, and rolling cathode and anode materials with binders and solvents to enhance adhesion and conductivity. The coated materials are then compressed into electrodes.

2. Cell Assembly: Integrating separators and electrolytes with electrodes, assembling them into mono-cells or bi-cells using stacking or winding techniques.

3. Formation Process: Charging and discharging cells to stabilize performance, followed by aging and degassing to remove trapped gases.

4. Pack Process: Configuring assembled cells into battery packs, connecting multiple cells, and enclosing them in protective casings for final use.

Project’s Social Impacts

The economic impact of NextStar Energy extends beyond its initial capital investment, serving as a catalyst for broader industrial transformation in Canada’s automotive sector.

As the Canada’s first large-scale EV battery plant, the project is shaping regional supply chains, workforce development, and economic resilience in ways that will endure long after its construction.

Job Creation and Workforce Development: With over 7,500 tradespeople engaged in its construction and 2,500 permanent manufacturing roles, NextStar Energy is a major employment driver. Workforce development programs, such as the Battery Boost initiative, are helping to train local talent, ensuring a steady pipeline of skilled workers for Canada’s growing EV industry (1). At its peak, the project is expected to involve 3,200 tradespeople and equipment installers, with 2,300 of those positions filled by Canadian workers (4).

Strengthening North America’s Supply Chains: Strategically located in Windsor—Canada’s automotive capital—the plant enhances supply chain efficiency by reducing dependence on imported materials. This integration strengthens North America’s EV ecosystem, making the region less vulnerable to external disruptions and creating a more resilient production network.

Attracting Ancillary Industries and Investment: The presence of a gigafactory of this scale is expected to attract key supporting industries, including raw material suppliers, logistics providers, and technology firms. This multiplier effect extends economic benefits beyond automotive manufacturing, reinforcing Windsor’s role as a critical node in the future of clean transportation.

The project has not been without controversy. The employment of 900 South Korean workers has drawn criticism from Canadian labor unions, who argue that domestic tradespeople are qualified for these roles. In response, NextStar Energy has defended the decision, stating that these workers are temporary specialists hired by external suppliers for highly technical installations essential to the plant’s operations (18).

Sustainable Development

EV Battery plants align with key United Nations Sustainable Development Goals (SDGs), integrating sustainability principles into its operations and setting a precedent for responsible industrial growth.

EV Battery plants affect the following four SDGs directly.

1. SDG 7 - Affordable and Clean Energy: By manufacturing batteries for EVs, the facility directly contributes to reducing greenhouse gas emissions and transitioning to renewable energy sources.

2. SDG 8 - Decent Work and Economic Growth: The project is expected to generate 2,500 jobs in manufacturing alone, alongside numerous indirect opportunities in construction, logistics, and supply chain management (24).

3. SDG 13 - Climate Action: The adoption of low-carbon energy solutions in battery production is aligned with Canada’s net-zero commitments.

4. SDG 12 - Responsible Consumption and Production: The integration of recycling processes ensures that raw materials are recovered and reused, minimizing environmental harm.

NextStar Energy is aiming to prove that economic growth and environmental responsibility can go hand in hand.

Windsor’s Automotive Legacy

Windsor’s emergence as an EV hub is the product of strategic planning, economic incentives, and cross-sector collaboration. Community Improvement Plans (CIPs) have played a key role, attracting private investment through financial incentives like property tax exemptions. The Brownfield Redevelopment CIP, for example, has delivered a 13:1 return on investment, underscoring the impact of public-sector incentives in driving economic growth.

After years of economic decline, Windsor is rebounding. Population growth (3% between 2011–2016) and a steady drop in unemployment signal recovery, while GDP growth, particularly in 2014–2015, marked a turning point. However, Windsor’s economy still lags slightly behind the national average.

Collaboration has been central to this revival. Invest WindsorEssex, local universities, and government agencies have worked to attract investment, develop training programs, and build specialized facilities like a state-of-the-art battery testing lab. Government support at all levels has anchored large-scale projects shaping Windsor’s industrial future.

Resilience has long defined Windsor’s automotive sector. From its early days as the "Border Cities" growing alongside Detroit, it has weathered downturns, bailouts, and trade shifts. The 1965 Auto Pact once secured investment in Canadian factories, but globalization, trade liberalization, and rising competition have eroded those advantages.

The NextStar battery plant is now at the center of Windsor’s economic reinvention. More than a manufacturing site, it’s a research and development hub for vehicle electrification, attracting further investment and solidifying Windsor’s role in advanced automotive technology.

Question then Becomes

The NextStar Energy project represents a landmark investment in Canada’s EV future, bringing together industrial ambition, government backing, and strategic foresight. Yet, the project exists at the intersection of technological change, policy shifts, and global competition—factors that will shape its trajectory in the years to come.

Nestled in Windsor, a city uniquely positioned to weather both supply chain vulnerabilities and regulatory headwinds, NextStar Energy appears well-placed to navigate the turbulence of the EV battery market. Yet, is this strategic positioning enough to secure the project’s economic success?

With the Trump administration proposing a 25% tariff on Canadian automotive components, the question then becomes: how will shifting trade policies and geopolitical uncertainties reshape the financial calculus of NextStar and similar EV battery ventures? Will the project’s economic foundation hold firm?

Notes

Join the EPM Network to access insights, influence our research, and connect with a community shaping the industry's future.

Subscribe now

Support us by sharing this article with your friends and colleagues.

If you wish to share your opinion, provide insights, or correct any details in this article, please email editor@epmresearch.com.

Refer to this article using the following citation format:

• Zangeneh, P., Abbasi M. (2025), “NextStar Energy EV Battery Plant: Canada’s First Advanced EV Battery Manufacturing Facility Megaproject” EPM Research Letters.

References

1. Yakub M. Electric Autonomy Canada. 2024. NextStar Energy begins battery module production in Windsor. Available from: https://electricautonomy.ca/ev-supply-chain/2024-10-23/nextstar-battery-windsor-production-canada/

2. Taylor I. Turnaround Cities: Anglo-Saxon Case Studies Insights from Pittsburgh (PA), Newcastle (NSW) and Windsor (Ont.).

3. Nextstar Energy. Official Website.

4. New NextStar Energy EV Battery Plant Celebrates ‘Topping Out’ Ceremony as Construction reaches 30 Per cent Completion. Stellantis Blog.

5. Windsorstar 2024. Windsor NextStar’s parent company reveals battery future plans.

6. Stellantis Media. NextStar Energy Announces Start of Battery Module Production.

7. Stellantis.com. Official Website.

8. LG Energy Solution. Official Website.

9. Alberici. NextStar Energy EV Battery Plant. Offical Website.

10. Kasian. EV Battery Plant: Powering Progress - Designing the Future of EV Batteries. Official Website.

11. Gresham Smith. Official Website.

12. Stantec to acquire Morrison Hershfield, a leading transportation, buildings, and environmental services firm. Press Release.

13. SMRT. NextStar Energy Electric Vehicle Battery Plant. Official Website.

14. Hodess Cleanrooms. Official Website.

15. Black & McDonald. Stellantis: Revolutionizing the North American EV Industry. Official Website.

16. Walters Group Inc. NextStar Energy EV Battery Plant. Official Website.

17. Alberici Hillsdale. Automotive | Hillsdale | Structural Steel Fabrication. Official Website.

18. Global News. Foreign workers at subsidized EV battery plant a ‘slap in the face’: union. Globalnews.ca.

19. CES Trump’s tariff threat spurs auto suppliers to rethink production plans. Reuters.

20. Mannes M. Exclusive: Scania tries to breathe life into troubled EV battery maker Northvolt. Reuters

21. Chu A, Smyth J, Temple-West P. $1bn US battery plant plan shows race to reduce reliance on China. Financial Times.

22. NextStar Energy Announcement on Pouch Batteries. LinkedIn Post.

23. LG Energy Solutions. ENSOLPEDIA | BATTERY INSIDE.

24. Automotive Logistics. Stellantis and LG Energy Solution begin battery production in Canada. Automotive Logistics.

If you’re in Montreal, let’s break bread and exchange ideas over lunch!

This is an opportunity to step away from Gantt charts, spreadsheets, risk matrices, and AI models to connect in person with others navigating the complexities of engineering project management.

📍 Location: TBD in Downtown Montreal, Please RSVP to editor@epmresearch.com
📅 Date: Friday February 21^st, 2025
⏰ Time: 12:00 to 1:30pm

Communities thrive on real-world conversations, and hopefully, this will be the first of many Canadian and international gatherings and networking events for our EPM community and network.

I look forward to meeting you, sharing updates on our current and upcoming research projects, and hearing your thoughts on them.

Regards

Pouya Zangeneh

Join the EPM Network to access insights, influence our research, and connect with a community shaping the industry's future.

Subscribe now

The IMEC study is quite unique. It was formed through the sponsorship of multiple stakeholders, who shared data and knowledge across different projects, that is then studied an interpreted by multiple scholars from various specialties. Therefore, minimizing host of biases in both data collection and result interpretation.

It is that rare instance where the proverbial elephant was well-identified in a well-lit room, where perceptive observers examined it together and discussed their perceptions too.

The Strategic Management of Large Engineering Projects by Roger Miller and Donald Lessard is a distillation of the outcomes of IMEC’s (International Program in the Management of Engineering and Construction) extensive study of sixty major projects across diverse sectors and geographies.

The IMEC program was commissioned in 1995 to study the strategic dimensions of large engineering projects (LEPs), and the book is published by MIT Press at the year 2000—25 years ago.

The program originated from the Hydro-Québec/CAE Chair in the Management of Technology at Université du Québec à Montréal and brought together academic institutions, industry leaders, and government agencies. The program was sponsored by organizations such as Atkin Realis (formerly SNC-Lavalin), Ontario Hydro, and the Project Management Institute.

The book, comprising ten chapters authored by different scholars, argues—at the beginning of the 21st century—that the foundation of project success lies in the early strategic design of institutional arrangements and governance structures.

In this review we cover:

1. What are Large Engineering Project (LEPs)

2. LEPs’ Main Strategic Dimensions

3. General Types of Project Sponsors

4. Risk Allocation by Project Financial Structuring

5. Strategizing to Face Risks and the Risk Layering Model

6. Project Shaping

   1. Strategizing Against Risks

   2. Iterative Process of Refinement

   3. Creation and Exercise of Options

7. Question then Becomes

The book explores the second-order strategic thinking, that is, deliberating actions that expand future response options in the face of unexpected events.

Several excellent chapters, such as Chapter 7 on building governability and Chapter 9 on partnerships, fall outside the limits this review. I encourage you to read the book firsthand, both for its extensive insights and as a thought experiment on how the management of Large Engineering Projects (LEPs) has evolved over the past 25 years.

What are Large Engineering Projects (LEPs)

The evolution in the meaning of engineering megaprojects and the complexities they face was recognized and highlighted by the authors in the early 2000s.

The authors defined Large Engineering Projects (LEPs) as massive, high-stakes ventures that shape economies and societies, and emphasize that the landscape of LEPs is shifting from centralized, government-led initiatives to complex partnerships involving private investors, regulators, and affected communities.

These projects must navigate regulatory hurdles, financial constraints, and political opposition while maintaining strategic flexibility.

The ability to shape institutions, forge alliances, and anticipate risks is what ultimately determines project success.

The book highlights six major hallmarks of LEPs and argues that by understanding these hallmarks, project leaders can shift from reactive management to proactive strategic positioning.

1. Product of Negotiated Compromise – LEPs evolve through complex negotiations among sponsors, investors, and regulators, requiring alignment of diverse interests and creative problem-solving.

2. Socially Contested Ventures – LEPs face political, environmental, and community challenges, with stakeholders competing for influence, making legitimacy and public approval critical.

3. Political Risks – LEPs are deeply intertwined with government policies, elections, and political shifts. Changes in leadership, national priorities, or public sentiment can drastically alter project viability.

4. Slow Development Cycles – Projects take decades to shape, finance, and execute, requiring patience and adaptability to shifting policies, economies, and market conditions.

5. Irreversible Investments – LEPs demand massive, non-recoverable capital commitments, making early-stage strategic planning crucial to avoid financial and operational deadlocks. Average investment at early 2000s timeline amounted to more than $1B.

6. Regulatory Complexity – Securing permits and approvals requires navigating layers of bureaucracy, often dealing with conflicting regulations across multiple agencies. Misalignment with regulatory frameworks can cause costly delays or outright cancellations.

LEPs’ Main Strategic Dimensions

In Chapter 1, the authors (Roger Miller and Donald Lessard) recount the main strategic dimensions and levers of LEPs.

They argue that successful Large Engineering Projects (LEPs) depend not just on technical expertise but on strong institutional arrangements. Institutional frameworks—laws, regulations, and governance structures—play a more crucial role in determining project performance than traditional planning and execution methods.

Projects anchored in stable and coherent institutions and predictable regulatory environments benefit from reduced risks, and clearer pathways for stakeholder engagement. In contrast, projects in weak institutional settings often struggle with instability, opportunism, and regulatory ambiguity, increasing their chances of failure.

Beyond institutions, the quality of sponsorship is a decisive factor in project success. Sponsors are responsible for shaping, financing, and managing risks across a project’s lifecycle. The authors posit that effective sponsors must master five essential skillsets:

1. Ownership Competency – Effective sponsors take full responsibility for project outcomes, ensuring strategic decisions align with long-term operational realities. They oversee engineering choices and contractual commitments to minimize future risks.

2. Strategic Evaluation – Sponsors must quickly assess project feasibility, sorting viable opportunities from high-risk ventures. This requires experience in financial modeling, regulatory analysis, and market forecasting.

3. Stakeholder Engagement – Building relationships with governments, regulators, and communities is crucial. Strong sponsors proactively manage public perception and political risks through transparent communication and negotiation.

4. Coalition-Building – Successful sponsors cultivate partnerships with investors, financial institutions, and industry specialists. These alliances help secure funding, manage risks, procure packages, construct project, operate them and navigate regulatory landscapes.

5. Resilience and Survivability – Sponsors must sustain long-term commitments, enduring financial setbacks and political shifts. A diversified project portfolio and strong financial backing increase their ability to adapt to crises.

General Types of Project Sponsors

The book identifies three major types of sponsors—network operators, concessionaire-developers, and ad hoc alliances. These sponsor types shape how LEPs are structured, funded, and executed, influencing their overall chances of success in volatile economic and regulatory environments.

1. Network Operators – Large, established firms that own and manage infrastructure networks, such as utilities or transportation systems. They bring financial stability and technical expertise but may struggle with bureaucratic inefficiencies.

   • Example: Endesa, a private utility in Chile, retained about 60% of the country’s power production even after deregulation. Their network ownership provided stability, but they had to navigate shifting regulatory landscapes.

2. Concessionaire-Developers – Private firms that specialize in financing, developing, and managing infrastructure projects, often operating across international markets. They are skilled at structuring financial deals and managing risks but can be vulnerable to political instability.

   • Example: INTERGEN, a joint venture between Bechtel and Pacific Gas & Electric, develops thermal power plants globally. Their expertise in risk allocation and international partnerships enables them to succeed in complex regulatory environments.

3. Ad Hoc Alliances – Temporary partnerships formed by entrepreneurs or engineering firms to develop specific projects. While agile and innovative, they often lack the financial backing and long-term credibility required for sustained success.

   • Example: The M1-M15 Toll Highway project between Vienna and Budapest was developed by an alliance between Transroute, Strabag, and European banks. The sponsors came together to finance and execute the project but had no long-term stake beyond its completion.

Risk Allocation by Project Financial Structuring

In Chapter 8, the authors (Richard Brealey, Ian Cooper, and Michel Habib) provide a coherent view of how financing large projects can help in dealing with risks.

Effective risk allocation in large infrastructure projects hinges on a delicate balance between control and risk-bearing capacity. Here, control refers to the ability of a party to directly influence, mitigate, or manage a specific risk through decisions, expertise, or operational actions.

For instance, contractors have control over construction risks because they oversee project execution, select materials, and manage timelines. Operators control operational risks through maintenance, efficiency improvements, and performance monitoring.

However, the capacity to control a risk does not always align with the ability to financially absorb its impact. Overexposure can lead to collapse when those managing the risks also bear excessive financial burdens. This creates a trade-off: fragmenting risk across multiple parties reduces individual exposure but weakens incentives for active risk management.

Some risks, such as currency fluctuations, cannot be controlled at the project level and are best transferred to financial markets. Others require a strategic alignment of ownership and responsibility so that parties both bear the consequences of risk and have the authority to mitigate it.

When direct risk allocation through contracts is impractical, financial architecture provides an alternative mechanism for distributing and renegotiating risks.

Unlike contracts that govern individual project aspects, financial structuring shapes control by determining who has decision-making power over financial resources or stands to gain or lose from the risk.

For instance, large banks or multilateral institutions with financial stakes in a project not only provide capital but also wield influence that can deter opportunistic renegotiation by governments or stakeholders, i.e., managing political risks. Since ownership confers control, financing structures, also, dictate who holds power in renegotiation scenarios.

Strategizing to Face Risks

In Chapter 3, the authors (Donald Lessard and Roger Miller) emphasize that risks in major projects are not isolated but interact dynamically, shifting in intensity over the project's life cycle.

The risk profile changes as the project evolve. Some risks, like regulatory approvals, diminish after initial hurdles are cleared, while others, such as financial and market risks, persist and evolve with economic conditions.

Technical uncertainties may be resolved after construction, but operational risks continue throughout a project’s lifespan.

Moreover, external shocks—macroeconomic instability, exchange rate fluctuations, or global financial crises—can compound multiple risk factors simultaneously. Institutional risks, particularly sovereign and regulatory uncertainties, can reignite financial and market vulnerabilities, making risk management an ongoing, adaptive process rather than a one-time mitigation effort.

These interdependencies highlight the need for robust contingency planning and flexible project structures to navigate uncertainty effectively.

The authors develop a taxonomy of risks in LEPs, categorizing them into market-related risks (market, financial, and supply risks), completion risks (technical, construction, and operational risks), and institutional risks (regulatory, social acceptability, and sovereign risks).

It is interesting to consider how certain risk categories, such as environmental risks, were not distinctly recognized as a separate entity in this risk taxonomy at early 2000s. Instead, they may have been subsumed under regulatory and social acceptability risks, reflecting the evolving understanding of sustainability and environmental concerns in project management.

The Risk Layering Model

Some risks can be anticipated and quantified, allowing sponsors to integrate mitigation strategies from the outset. Others, however, arise dynamically—triggered by political shifts, stakeholder opposition, or macroeconomic turbulence—forcing project sponsors to adapt and renegotiate their strategies in real time.

The key challenge is not merely to identify risks but to develop strategic responses that align risks with the capacities of different stakeholders.

Authors argue that traditional risk models, which rely on probabilistic forecasting, often fall short in capturing the complex interdependencies that shape project outcomes. Instead, successful sponsors approach risk as an active management challenge, layering their strategies to create resilience, adaptability, and long-term viability.

To navigate this uncertain terrain, project sponsors must layer their risk strategies, balancing direct risk allocation through contracts with indirect risk management through financial and governance structures.

The layering model breaks down risk management into six core strategies: (1) assessing and understanding risk, (2) shifting risks through financial instruments, (3) diversifying risk exposure, (4) creating options for adaptive responses, (5) transforming risks through institutional engagement, and (6) bearing residual risks as a calculated choice. Risks are not simply avoided but strategically assigned, diversified, transformed, or embraced, depending on their nature and controllability.

Figure 1 illustrates these six layers of strategies for managing risks. The book elaborates on each of these strategies in valuable details. By integrating these layers, project sponsors do not merely react to risk—they shape project trajectories, ensuring that risks are embedded within a structured framework that enhances governability and long-term project success.

Project Shaping

In Chapter 4, the authors (Roger Miller and Xavier Olleros) define project shaping as an active and iterative process where sponsors do not merely react to uncertainties but strategically engage with risks to influence outcomes, that unfolds over multiple episodes, typically lasting six to seven years.

Successful shaping requires expertise in organizational strategy, law, diplomacy, and public affairs. Sponsors must navigate complexity by forming alliances and coalitions, leveraging prior experience, and incorporating expert advice. This approach transforms significant uncertainties into manageable challenges by distributing responsibilities and aligning incentives across stakeholders.

Sponsors start with a configuration hypothesis, testing and adjusting it through strategic interventions. Each shaping episode addresses specific challenges, such as market uncertainties, regulatory hurdles, or stakeholder opposition, leading to incremental refinements.

This procedural rationality—where adaptation takes precedence over rigid planning—allows projects to evolve dynamically, ensuring that unforeseen disruptions do not derail long-term goals.

The authors make an interesting observation: effective shaping means balancing control with flexibility, ensuring that projects remain adaptable while maintaining strategic momentum. By continuously refining their approach through successive shaping episodes, sponsors gradually bring projects to full momentum and closure, creating structures that are resilient, governable, and primed for long-term success.

Project shaping arises from the mindset of not avoiding complexity but embracing it when it is worth doing so. Sponsors must manage nonlinear interactions, interdependencies, and shifting institutional landscapes, making strategic decisions that account for emergent risks.

Chapter 4 sets the book’s narrative in developing strategic moves to anticipate uncertainty and risks through out the project within the context of project shaping.

The idea of project shaping is discussed through the book in three main contexts of Strategizing Against Risks, Iterative Process of Refinement, and Creation and Exercise of Options, albeit, in different chapters. Here is my attempt to summarize them accordingly.

Project Shaping as Strategizing Against Risks

Project shaping as strategizing against risks integrates risk management into the foundation of large engineering projects (LEPs), ensuring resilience and adaptability. Effective sponsors use risk-layering strategies—assessing, shifting, diversifying, creating options, transforming, and bearing risks—to proactively align projects with emerging uncertainties. By engaging stakeholders early, they identify disruptions, align incentives, and refine project configurations to minimize failure risks.

Risk layering distributes uncertainties strategically, enhancing project adaptability. Assessing risks ensures informed decision-making, while shifting and diversifying risks limit exposure. Creating options, such as modular investments, builds flexibility, and transforming risks through regulatory influence secures legitimacy. Bearing residual risks maintains control, allowing sponsors to influence outcomes effectively.

By applying these strategies, sponsors shape projects into robust, governable entities rather than leaving them vulnerable to uncertainty. Successful LEPs engage with risk strategically, using uncertainty as a tool for coalition-building, financial stability, and long-term viability. Project shaping, therefore, is less about avoiding risks and more about structuring uncertainty to create sustainable, high-value ventures.

Project Shaping as an Iterative Process of Refinement

Project shaping is not a singular decision but an evolving sequence of progressive refinements, where sponsors navigate multiple shaping episodes to align technical feasibility, financial viability, regulatory compliance, and stakeholder expectations. Each episode serves as a hurdle, requiring adaptation and negotiation to maintain project momentum.

Figure 2 depicts a decision tree for navigating project hurdles over multiple episodes.

Sponsors begin with broad conceptual assumptions and, through successive cycles of coalition-building and risk assessment, refine their approach to meet emerging challenges.

This process is dynamic, often facing early lock-in risks that can restrict future flexibility or require renegotiations when unforeseen issues arise. Effective shaping involves continuous stakeholder engagement, ensuring that governance structures, financial agreements, and technical designs remain adaptable while balancing competing interests.

In doing so, shaping becomes a form of strategic maneuvering—aligning incentives, mitigating risks, and integrating social and environmental considerations to transform uncertainty into governable structures.

Project Shaping as Creation and Exercise of Options

Beyond iterative refinement, shaping is a mechanism for creating optionality, allowing sponsors to respond dynamically to changing conditions rather than committing prematurely to fixed solutions.

The real-options framework, drawn from financial theory, is useful to conceptualize how uncertainty can enhance project value if structured decisions are made sequentially rather than in a single, irreversible move.

By keeping strategic choices open—whether in financing, technology selection, or stakeholder agreements—sponsors can delay critical commitments until more information is available, reducing downside risks while maximizing upside potential.

This approach demands flexible governance, where partnerships are structured to support adaptability rather than rigid control. The authors recount a few examples to clarify this idea. Premature lock-in, such as the early drilling technology choice for the English Channel Tunnel, demonstrates the risks of closing options too soon, while the Brazilian ITA project illustrates how deferring financial decisions until the right conditions emerge can lead to a more sustainable outcome.

The key to financial shaping lies in maintaining the ability to recalibrate commitments—whether through staged investments, contingent financing agreements, or leveraging contractual safeguards—ensuring that projects not only secure initial funding but remain resilient against evolving economic and political landscapes.

Question then Becomes

Nearly 25 years after the IMEC research was published, its insights remain foundational, demonstrating remarkable foresight in anticipating the evolving landscape of LEPs. The study not only captured the complexities of its time but also laid out the direction of more than two decades of industry transformation.

At the same time, the evolving landscape of large engineering projects demands adaptation. Sustainability mandates, digital transformation, and geopolitical shifts are reshaping project governance, risk management, and financial structuring.

The rise of green bonds and impact investing calls for new financial models that balance economic viability with environmental imperatives. Meanwhile, stakeholder dynamics are more complex than ever.

What are the most important dimensions in which IMEC’s strategic framework must evolve to stay relevant? How can lessons from IMEC guide energy transition projects in navigating regulatory hurdles, community engagement, and Indigenous partnerships?

The genius of the IMEC program lies not only in its findings but in its very existence—the fact that such a comprehensive, multi-stakeholder collaboration actually took place.

The book leaves the reader contemplating whether a similar large-scale study could be conducted today and, if so, what its outcomes might be.

I cannot help but wonder: what conditions would be necessary to foster a similarly ambitious initiative today?

Notes

This book review is a reading material for the “Strategy in Project Management: Game Theoretic Approach,” course at the University of Calgary, Department of Civil Engineering.

Refer to this article as below:

• Zangeneh, P. (2025), “The Strategic Management of Large Engineering Projects: Shaping Institutions, Risks, and Governance by Roger Miller and Donald Lessard,” EPM Research Letters.

Join the EPM Network to access insights, influence our research, and connect with a community shaping the industry's future.

Subscribe now

In this second part, we journey deeper into the mechanics of strategy, examining the tools and techniques that make strategies credible and interactions effective. We will navigate along with the book to review the following subjects.

• Making Strategies Credible,

• Information Interpretation and Manipulation,

• Cooperation and Coordination,

• Bargaining and Negotiations,

• Voting, and,

• Incentives.

As in the first part, I illustrate the book’s principles by examining their application through the lens of a project manager navigating an urban light rail megaproject. This context, I hope, demonstrates how game theory serves as a compass in the often uncharted terrain of project management.

I hope this review inspires you to engage with the book itself. The Art of Strategy is the result of distilling several books through their writers' journeys into its finest version. As they rightly advertised, it serves as a guide for success in business, life, and, of course, project management.

Making Strategies Credible

Building on the previous section, strategic moves consist of two critical components: the move itself and the course of action that establishes its credibility.

Credibility is pivotal because even the boldest moves lose their impact if the opposing player doubts your ability or willingness to act.

As discussed earlier, when faced with threats, a sound strategy is to undermine their credibility, such as exploiting weaknesses in the opponent’s commitment mechanisms.

Another approach is to test and incrementally challenge the credibility of a threat through a concept famously theorized by Thomas Schelling as salami tactics. This involves taking a series of small, calculated steps that gradually push the limits of an opponent’s commitment. Each move is designed to appear too minor to warrant a disproportionate response, making retaliation seem excessive or unjustifiable. Over time, these seemingly inconsequential actions accumulate, effectively weakening the opponent's original threat and shifting the strategic balance.

Salami tactics enable one side to advance its position incrementally while forcing the other to either escalate dramatically or concede ground. By exploiting the opponent's hesitation to act decisively, this approach exposes the fragility of threats that lack robust and enforceable credibility.

On the other hand, when confronted with salami tactics, you might restrict your own choices by cutting off communication or delegating decision-making to an automated system. If you clearly communicate the existence of such a mechanism, this strategy relieves you of the burden of acting proportionally and strengthens your position by removing room for hesitation.

Dixit and Nalebuff classified the approaches to establish credibility for strategic moves into four category of tactics:

1. Changing Payoffs: Restructure the game to make compliance more attractive or deviation prohibitively costly. This includes writing enforceable contracts, and establishing and leveraging reputation.

2. Restricting Choices: Constrain the ability to reverse commitments by cutting off communication to limit renegotiation, or, burning bridges, such as removing fallback options to create an all-or-nothing scenario, or, delegating outcomes to chance or automated processes.

3. Incremental Moves: Break large commitments into smaller, manageable steps, fostering trust while reducing risks. This tactic also addresses the salami strategy by ensuring that each incremental action can still enforce discipline or resolve.

4. Using Agents and Teams: Enhance credibility through external enforcement such as mandated agents, or social and team pressure.

Automatic Escalation

As a project manager, you are often dealing with contractors struggling with delays and under high stress. Imagine in our urban light rail projects, the contractors frequently fail to submit intermittent progress commitments and often provide excuses instead of actionable plans, making it increasingly likely that the final deadline will be missed.

To address this, you may leverage the risk management function to create an automated mechanism that escalates delays to the project's higher management as material risks. These risks are configured to trigger when any contractors miss their specific milestones or fail to submit required documentation, such as updated work schedules or mitigation plans. Once triggered, the risk is automatically flagged in the project’s centralized dashboard, making it visible to all stakeholders, including higher management.

The system restricts your ability to personally intervene or renegotiate these escalations, effectively restricting your choices and creating a scenario where the contractors face the consequences of their delays without the usual mediation or leniency. This strategy removes fallback options for the contractors and signals the seriousness and the credibility your strategy.

Information Interpretation and Manipulation

In real-world strategic situations, it is common for one player or party to possess more information than the other, creating an imbalance in information, or, information asymmetry. This asymmetry—or even the perception of it—can be leveraged or manipulated to influence outcomes in favor of the more informed party.

To navigate these dynamics, mechanisms like signaling, screening, and signal jamming play a role in shaping how information is conveyed, interpreted, and controlled.

Signaling involves a player demonstrating attributes or intentions through observable actions, often to influence others’ decisions. For a signal to be credible, it must carry a cost or inconvenience that deters false representation. Cleanliness, for instance, is not a reliable signal of quality of a used car, as its relatively cheap to have any car cleaned. In contrast, car dealerships or companies may use signals such warranties to affirm product quality—implicitly showcasing their confidence by bearing the potential cost of defects.

Screening, unlike signaling, is initiated by the less-informed party to elicit private information from others. It involves designing scenarios where different actions reveal differences in underlying qualities or risk profiles. Devising screening mechanisms could lead to unintended effects. In insurance markets, for instance, policies with uniform pricing may attract disproportionately high-risk individuals while discouraging low-risk participants, this is called adverse selection. This is why screening techniques in such industries typically aim to predict individuals risk levels indirectly through their seemingly unrelated choices.

Finally, signal jamming is the strategic obfuscation of information to prevent others from accurately discerning one’s intentions. This tactic is common in competitive or adversarial contexts, where concealing true motives or capabilities can confer an advantage. Poker players employ reverse tells to confuse opponents. Many counter espionage operations such as Canary Traps, Honey Traps, or False Flags, are built based on deceptive signals to deliberately mislead.

Signal Jamming in Projects

Teams on a large transit project might engage in signal jamming by strategically repackaging their deliverables to highlight easy wins while masking the root causes of significant issues.

For instance, they could emphasize timely completion of auxiliary tasks, such as routine designs or minor installations, in progress reports while underreporting delays or overruns in core components like tunnel construction or station infrastructure.

This selective reporting shifts attention to areas of perceived success, deflecting scrutiny from critical challenges. By focusing on partial deliverables or separating problematic areas from successful ones in updates, teams create a misleading narrative that obscures the true status of the project. This makes it harder, if not impossible, for management to identify and address root causes of delays or budget overruns.

Earned Value Management (EVM) methods, while valuable for tracking project performance, are particularly susceptible to such signal jamming. Teams can inflate the earned value by showing progress on less critical, lower-effort tasks that contribute little to the overall project while neglecting high-impact activities facing delays or cost issues. Progressively, this will force teams to cut corners on other parts of the project to avoid exceeding budget or schedule constraints, leading to compromised quality. For example, materials or workmanship might be downgraded on less visible project elements to counterbalance overruns elsewhere, resulting in long-term operability issues.

Implementing proper incentive schemes, transparency measures, and commitment mechanisms can counteract these issues and discourage signal jamming. Review and updating the rules of credit for EVM analysis is of paramount importance.

Signal jamming happens on most large projects and to various extents because of piecemeal decision-making as discussed in the following section.

Cooperation and Coordination

As outlined in the first part of this review, individual decisions can collectively lead to suboptimal outcomes. And, there is no way to remedy that other than by changing the payoffs.

Success, in many domains, is a game of relativity rather than absolutes. Just as OPEC nations coordinate production limits to prevent a destructive price war, so too do industries. Over-competition, such as excessive studying for grades or excessive training in sports, leads to a collectively suboptimal outcome. True strategy, then, is not just about self-improvement but about understanding when to compete, when to cooperate, and when to redefine the game entirely. This can be thought as the inverse of the Prisoners’ Dilemma, when collective agreements (from Unions to Cartels) is often the winning strategy.

In many systems, small shifts in dynamic behaviors or conditions may cascade into significant, systemic shifts. This phenomenon is known as a tipping point—a critical threshold beyond which collective behavior changes dramatically, often locking the system into a new—extreme—equilibrium. For instance, in racially mixed neighborhoods, minor demographic shifts can set off self-reinforcing dynamics that result in complete segregation. Overcoming these tipping points requires deliberate, coordinated efforts to realign the system toward more favorable equilibria.

These small shifts can be agreed upon by all parties along the way. Often seemingly rational incremental steps, through piecemeal decision-making, can lead to unintended, collectively regrettable outcomes.

Imagine a system where individuals collectively determine promotion criteria, starting from a low standard and gradually raising it in small steps. Each decision appears logical to the majority: high-achieving members benefit from enhancing the group's perceived quality, while those who are excluded prefer a higher standard to blur their relative ranking among the rejected.

Over time, these incremental steps escalate to an extreme threshold where no one achieves promotion, leaving everyone worse off. Despite this outcome, each decision along the way was supported by most participants, illustrating how localized, self-serving choices can fail to account for their cumulative, detrimental effect on the group. This phenomenon is called of piecemeal decision-making or incremental rationality

Incremental Rationality and Piecemeal Decision-Making in Projects

Incremental Rationality is often mirrored in large projects with multiple sub-projects, where incremental change orders and scope adjustments occur. For example, within a construction megaproject, sub-project teams may propose minor design changes or additional features that seem beneficial and manageable in isolation.

Each change order, approved by the majority of project leadership and stakeholders, shifts the project slightly from its original scope. Over time, however, the cumulative effect of these changes can lead to significant budget overruns and delays, diverting focus from the project's core objectives. As these incremental decisions pile up, the project as a whole begins to lose alignment with its initial priorities, often with the benevolent approval or consent of all involved parties.

This collective approval sets the stage for signal jamming, shifting the narrative away from addressing systemic issues, creating a self-reinforcing loop where the group avoids acknowledging failures and instead justifies its actions through selective reporting or framing.

Avoiding piecemeal decision-making is critical, and may start with bundling potential changes (change orders) into comprehensive packages to allow stakeholders to evaluate their cumulative impacts—with transparency and accountability—rather than approving them individually.

Bargaining and Negotiations

Bargaining is a careful balance of strategy, where each side pushes for its interests while competing for shared resources.

Whether in scenarios like wage negotiations or contractual disputes, outcomes are shaped by critical factors such as the cost of delay, external alternatives, and the distribution of bargaining power.

Bargaining can be viewed as a sequential game of offers and revisions, where negotiators benefit from backward reasoning to secure agreements that maximize mutual gains while minimizing avoidable losses, such as those arising from strikes or prolonged disputes.

A central concept in bargaining and negotiation is BATNA (Best Alternative to a Negotiated Agreement), which defines each party’s fallback position if talks fail. BATNA serves as a baseline for determining the value each side can secure independently, heavily influencing the division of the "bargaining pie."

Successful negotiations hinge on creating value beyond the sum of the BATNAs, with the surplus divided according to each party's relative bargaining power. Strengthening one’s BATNA—or weakening the opponent’s—is a key strategy for gaining leverage in bargaining and negotiations. Effective negotiation requires precise assessment of BATNAs, anticipation of the other party’s actions, and creativity in expanding the pie to achieve outcomes that benefit all involved.

The Rubinstein Bargaining model offers a foundational framework for understanding these dynamics. This model examines negotiations over a divisible resource through alternating offers, assuming complete information for all parties. The model representation demonstrate that bargaining power is linked to the negotiators' patience. Patience, quantified as a discount factor (δ), indicates how much a party values immediate resolution versus future outcomes.

Greater patience in negotiations allows a party to hold out longer, increasing their chances of securing a larger share of the resource. In contrast, those with less patience feel more pressure to settle quickly, often making greater concessions to reach an agreement sooner.

Bargaining in Project Disputes

Consider a scenario where the delivery of train wagons for a large public transit project is delayed by the supplier.

The delay has resulted in three types of costs: the contractor’s direct costs (e.g., idle labor and machinery), shared mitigation costs (e.g., sourcing alternatives or accelerating work), and the owner’s indirect costs (e.g., reputational damage, penalties, and lost revenue due to delays).

The critical question becomes: who is more impatient in this negotiation—the contractor or the owner? The contractor’s impatience is shaped by their financial exposure, such as cash flow pressures and resource constraints. However, if the contractor has operational flexibility, such as reallocating idle resources to other projects their level of impatience may be reduced. Maintaining a healthy portfolio of active projects increases the patience of contractors in bargaining disputes.

On the other hand, the owner typically faces higher indirect costs, including public scrutiny, political commitments, and potential penalties, all of which create a stronger urgency for resolution. If the owner ends up more impatient than the contractor, bargaining dynamics will results with the owner bearing the majority of the shared mitigation costs.

Strategies to remedy the bargaining imbalance are crucial for project owners. The owner can strengthen their BATNA (Best Alternative to a Negotiated Agreement) by introducing penalty clauses in contracts that shift more risk to suppliers or contractors. But aside from contractual arrangement, the owners should diversify their supplier options to increase redundancy.

Voting

Voting, in any system, is inherently strategic.

Voters often realize that simply choosing their favorite candidate may not always lead to their desired outcome. Instead, they must consider how others will vote and whether supporting other options will better serve their interests.

This strategic dynamic becomes especially pronounced in elections with three or more candidates, where casting a ballot for one’s true preference can inadvertently lead to an undesirable result. Voters may choose to support a candidate who is not their top choice but stands the best chance of defeating an unfavorable alternative.

To address these challenges, ranked-choice voting has been proposed as a means to relatively mitigate strategic voting. This system allows voters to rank candidates in order of preference, enabling a more nuanced expression of voter sentiment. It effectively simulates runoff elections within a single voting round, reducing the risks of vote-splitting and wasted votes. Approaches like the Condorcet method—where the winner is the candidate who would prevail in all head-to-head matchups—are particularly robust but require ranked ballots to be practical.

In a one-dimensional political spectrum, candidates tend to gravitate toward the preferences of the median voter, as this position secures majority support. This convergence incentivizes truthful voting, as voters have no reason to misrepresent their preferences when outcomes hinge on the median. While effective in simpler, one-dimensional settings, the median voter theory becomes less stable in multi-dimensional contexts where issues are more complex and diverse. Nevertheless, the stability of median voter outcomes is a cornerstone of political and organizational systems.

Strategic Proposals

While voting may seem distant from project management, the principles of strategic alignment with stakeholder preferences—such as those in median voter theory—are directly relevant when crafting proposals for evaluation by decision-making bodies like private or governmental committees. These committees often represent diverse interests, weighing dimensions such as cost-efficiency, environmental sustainability, long-term economic benefits, regional equity, and public support.

Understanding and mapping these priorities allows for the identification of the "median position"—a balanced point that reflects the committee’s central concerns. Structuring a proposal to align with this median position maximizes its appeal.

To address fringe preferences without compromising the core strategy, thoughtful amendments can be included. For example, if a subset of stakeholders strongly advocates for more aggressive environmental measures, the proposal might integrate optional green enhancements, such as additional scope for renewable energy initiatives.

Beyond potential scoring systems for evaluating alignment with stakeholder priorities, factors such as credibility, technical rigor, the overarching direction advocated by a proposal play a pivotal role. A core consideration is ensuring that the proposal's priorities align seamlessly with the strategic objectives of the project owner organization. Nonetheless, strategically centering the proposal on the median position—reflecting the most balanced and widely shared priorities—while thoughtfully integrating concessions for outlier positions creates a compelling and adaptive approach.

Incentives

Incentives are another key mechanism for altering strategic situations and aligning individual efforts with desired outcomes. When designed thoughtfully, they can bridge the gap between individual motivations and organizational goals, particularly in contexts where tasks have a probabilistic nature.

Effective incentive systems balance fixed rewards with performance-based rewards, recognizing the imperfect relationship between effort and outcomes. For tasks influenced by factors beyond an individual’s control—such as sales or project-driven work—this balance is critical. A mix of base pay and bonuses tied to observable results encourages effort without unfairly penalizing for uncontrollable variables.

However, calibrating these incentives is a delicate task: overemphasis on performance-based rewards in uncertain environments can demoralize workers, while relying too heavily on fixed compensation risks dampening motivation. Tailored incentive schemes drive engagement and productivity by aligning rewards with the specific context of the work.

When tasks are interconnected, whether as substitutes or complements, incentives must be designed with these relationships in mind. Substitutable tasks, like balancing teaching and research, can compete for a pool of attention. If incentives heavily favor one over the other, it may lead to neglect of the less-rewarded task, requiring careful balance to avoid unintended trade-offs. Complementary tasks, such as beekeeping and orchard management, benefit from well-aligned incentives that reinforce their natural synergy, leading to greater overall productivity.

Requiring university professors to teach a research course in their research area is therefore a powerful strategy to align their incentives.

This logic also applies to organizational structure: grouping complementary tasks together fosters coherence and efficiency, while separating substitutable tasks helps reduce internal competition and misalignment.

Incentives for Project Managers

In project management, the role of luck cannot be overstated. From unforeseen weather disruptions to supply chain delays, numerous factors beyond the control of teams can significantly impact progress. Designing proper incentives in this context requires moving beyond a singular focus on project completion.

Overemphasis on final delivery often leads to compromised quality, safety shortcuts, and critical issues being hidden to meet deadlines. To counter this, incentives should be progress-adjusted for context, ensuring that teams are rewarded for effort and adaptability rather than raw outcomes alone.

Performance metrics should be carefully chosen to reflect achievements distant from the probabilistic nature of risk events, such as compliance with intermediate quality benchmarks, proactive risk mitigation, or maintaining team morale in challenging conditions. This approach fosters accountability and fairness, encouraging teams to perform at their best despite external uncertainties.

Equally important is addressing the inherent conflict between substitute goals such as quality, safety, and productivity.

When these achievements are at odds—such as prioritizing speed at the expense of safety—aligning incentives becomes a challenge. Separating rewards and even departmental responsibilities for these objectives can reduce misalignment. For instance, a safety team could be tasked solely with enforcing safety standards and rewarded based on incident-free days, while a productivity team focuses on delivering milestones on schedule. By decoupling these incentives and responsibilities, teams can excel in their areas without undermining the others. This structured approach not only minimizes counterproductive competition but also ensures that core project values like safety, quality, and efficiency are upheld in balance.

Question then Becomes

As we navigate the complexities of strategic decision-making, several critical questions emerge.

Are there ethical concerns with using commitment mechanisms, such as delegating decisions to automation, to establish credibility? As AI continues to evolve, this issue will become increasingly relevant, raising questions about accountability, fairness, and unintended consequences.

Similarly, how can organizations prevent or minimize signal jamming, especially in large projects where selective reporting is common? When teams manipulate information to highlight minor successes while concealing critical delays, decision-makers lose visibility into the true status of a project.

Politicians often capitalize on the intuitive instincts of their supporters, shaping narratives that resonate deeply. For instance, Donald Trump frequently emphasized the unpredictability of his strategies on certain topics, signaling—whether explicitly or subliminally—that his approach was designed to ensure victory for his party, often at the expense of opponents. Simultaneously, by projecting an air of absolute honesty and transparency, he conveyed the idea of a broader, inclusive victory, suggesting that everyone stood to gain. This dual signaling taps into the natural dichotomy of zero-sum and non-zero sum game dynamics in both his supporters and his enemies. Can you think of instances where he has employed this strategy?

These questions highlight the intricacies of applying game theory principles to real-world scenarios. Who else in your friends or colleagues would find these ideas useful?

Notes

This book review is a reading material for the “Strategy in Project Management: Game Theoretic Approach,” course at the University of Calgary, Department of Civil Engineering.

Refer to this article as below:

• Zangeneh, P. (2025), “The Art of Strategy - Part 2: A Review of Dixit and Nalebuff’s Classic Within the Context of Project Management”, EPM Research Letters.

Join the EPM Network to access insights, influence our research, and connect with a community shaping the industry's future.

Subscribe now

Potash, primarily used as a key ingredient in fertilizers, enhances crop yields, strengthens plant resistance to pests and diseases, and improves drought tolerance. With a growing global population and food demand projected to double by 2050, Canada's leadership in potash production is more vital than ever.

Most of Canada’s potash reserves are in the province of Saskatchewan. This region, hosting some of the richest Potash deposits on the planet, is home to ten active mines and several major development projects. In 2022 alone, these mines produced approximately 24.6 million tonnes of potassium chloride (KCl). The province's extensive rail and export infrastructure further amplify its role in feeding the world.

The potash industry in Canada is a sponsor of many large engineering megaprojects, such as BHP’s Jansen Potash Project, one of the largest capital investments in the sector globally. These projects are engineering marvels and underscore the industry's role as a catalyst for innovation, sustainability, and growth within the heavy industry construction sector of Canada.

In this article we review,

1. Potash and Potash Mining Methods

2. History of Canada’s Potash Industry and its Highlights

3. Saskatchewan Potash Reserves and Operating Mines

4. Canada’s Potash Transportation System

5. Capex Requirement of Potash Megaprojects

6. The Curious Relationship of Potash and Climate Change Adaptation

From Calgary, to Saskatoon, to Winnipeg, we embarked on a 14-hour long road trip to witness the heart of Canada's agricultural legacy, not to mention the vibrant city of Saskatoon. This road trip left us with a much deeper appreciation of Canada’s role in feeding the world.

Voyage and Return

Last year, our research group planned a trip to Winnipeg for a joint conference. The idea of turning it into a road trip through Canada’s potash heartland seemed exciting at first. But once we realized it involves a 14-hour drive, our enthusiasm quickly faded away.

As we dug deeper into Canada’s rich potash legacy, the road trip idea grew on us. Our most viable plan was to rent a car, leave Calgary at dawn, and to fly back from Winnipeg. After weighing the pros and cons, we decided to put it to a group vote. Unanimous, the road trip won.

What followed was an unforgettable journey through some of the world’s largest potash mines and the impressive transportation networks that make this industry possible. A proud legacy for Canada. Looking back, it was definitely a long trip, but the experiences and insights we gained, and sharing them with you, will make it worthwhile.

Potash

Potash, derived from the Dutch "potasch" meaning "pot ash," is used to describe a variety of potassium-bearing minerals and compounds.

This term originates from the traditional method of extracting potassium carbonate by leaching wood ashes in pots. Today, "potash" primarily refers to potassium chloride (KCl), but also includes potassium sulfate (K2SO4) and potassium nitrate (KNO3), all crucial for manufacturing fertilizers.

Potash is a key ingredient in fertilizers that improve crop yields, enhance plant resistance to diseases and pests, and increase drought tolerance. Potassium is vital for plant health, aiding in functions like photosynthesis, water regulation, and disease resistance. These benefits make it indispensable in modern farming practices, especially as the global population and food demand continue to rise.

The limited availability of Potash poses significant challenges to agricultural growth. With the global population rising and food demand expected to double by 2050, potash has becoming an increasingly vital mineral (1). The mining of potash involves extracting these potassium-rich minerals from underground ore deposits, which are then processed into usable forms for agriculture (2).

Before the pandemic in 2021, potash prices comfortably sat between CA$250 and 280 per metric ton. But as the pandemic unfolded, the disruptions in supply chains, heightened demand for agricultural inputs, and geopolitical pressures caused prices to soar to more than five times the pre-pandemic levels, reaching over CA$1,500 per metric ton in early 2022.

Today, prices have begun to stabilize, settling around CA$400 to 450 per metric ton, about 1.5 to 2 times pre-pandemic levels.

Potash Mining

There are two primary methods for potash mining: conventional underground mining and solution mining.

Conventional underground mining involves excavating deep shafts to access deposits at depths of 900 to 1,100 meters, where machinery and workers extract the ore. This is most suitable for shallow, high-grade deposits and is widely used across seven mines in Saskatchewan.

In contrast, solution mining, used at three mines, targets deeper deposits up to 1,500 meters, or is used to extract from disturbed or abandoned former underground mines, of instance due to previous flooding. This less invasive technique involves injecting water to dissolve potash into a brine solution, which is then pumped to the surface (4).

We will compare the capital cost requirements of these methods further in this article.

The processing of potash includes refining the material to remove impurities and increase the concentration of potassium chloride through techniques such as flotation, crystallization, and drying.

Canada’s Potash Industry

Economically, potash is a cornerstone of Canada’s resource sector. The industry contributes 23 billion of dollars to Canada annually, supports 76000 jobs (5), and contributes to the economic development of rural and Indigenous communities. Canada’s extensive transportation infrastructure, including railways and ports, facilitates the efficient export of potash, ensuring its accessibility to global markets.

History

• 1940s: Early Inspiration and Exploration – Potash exploration in Canada began, inspired by discoveries in Germany and the United States, driven by rising global demand for agricultural fertilizers.

• 1943: Accidental Discovery – Potash was discovered in Saskatchewan during routine oil drilling, but its significance wasn’t immediately recognized due to low market demand and focus on oil.

• 1950s: Rising Demand and Focused Surveys – Global demand for potash grew, prompting geological surveys in Saskatchewan that revealed extensive deposits suitable for mining.

• 1958: Birth of Canada’s Potash Industry – The first potash mine near Saskatoon was opened by the Potash Company of America, officially starting Canada’s potash production.

• 1960s: Expansion Era – Potash demand surged, leading to the opening of several new mines in Saskatchewan and a significant increase in production capacity.

• 1970s: Oversupply and Industry Challenges – A global surplus of potash caused price drops and reduced production, prompting industry consolidation and technological advancements for efficiency.

• 1980s–1990s: Stabilization and Innovation – The industry stabilized with moderate growth driven by innovations in mining and processing, improving both output and environmental sustainability.

• 2000s Onward: Renewed Expansion – Rising global food demands led to increased production and expansion, with companies like PotashCorp (now Nutrien) playing a major role.

• Today: Saskatchewan’s Global Leadership – Saskatchewan is the largest potash-producing region worldwide, holding nearly half of the global reserves.

Innovation

The Canadian potash industry has a legacy of innovation that has propelled it to global leadership.

Early mining efforts faced significant geological obstacles, including severe water inflows during shaft sinking. To address this, ground-freezing techniques were pioneered, enabling safe and stable underground mining. These breakthroughs laid the foundation for Canada’s potash sector to flourish (6).

In 1964, Kalium Chemicals Ltd. opened the world's first potash solution mine near Regina. The introduction of solution mining allowed access to reserve unreachable by conventional mining techniques. It not only expanded production capabilities but also demonstrated the industry's adaptability to complex geological conditions (7).

In the 2000s with the Saskatchewan Research Council (SRC) developed new potash assay methods, becoming one of the world's largest potash assay laboratories. This supported the industry's growth by providing accurate analysis for exploration and production (8).

Today, the potash industry continues to lead with innovations focused on sustainability and efficiency. Practices such as water recycling, land reclamation, and digital optimization are now integral to mining operations. This enduring commitment to innovation ensures that Canada remains at the forefront of the global potash market while contributing to more sustainable agricultural practices worldwide (9).

Social and Economic Impacts

The potash industry is a vital pillar of Saskatchewan's economy, significantly contributing to both the province and Canada as a whole.

In 2022, it added approximately CA$8.6 billion to Saskatchewan’s GDP and generated CA$5.4 billion in tax revenues across federal, provincial, and local governments. This financial impact highlights its critical role in sustaining public services and supporting economic growth.

Over the past 15 years, Saskatchewan has attracted CA$30 billion in investments for new mines and expansions, showcasing the industry's dynamic growth. On a national level, Canada’s fertilizer industry, anchored by potash, contributes roughly CA$23 billion annually to the economy and supports over 76,000 jobs across the country. As demand for agricultural fertilizers continues to grow globally, the Canadian potash industry remains a cornerstone of economic stability and innovation (10, 11)

Saskatchewan Potash Reserves and Operating Mines

Saskatchewan is a province in the heartland of Canada known for its vast horizons and rich agricultural land.

It is bordered by Alberta to the west, Manitoba to the east, and the Northwest Territories to the north. Saskatchewan is famous for its diverse landscapes, which include rolling plains, dense boreal forests, and numerous lakes and rivers.

The province is a major producer of grains and oilseeds, making agriculture a cornerstone of its economy. Saskatchewan is rich in minerals, particularly potash and uranium, and is a leader in energy production from oil, natural gas, and renewable methods. Its major cities, Regina, the capital, and Saskatoon, the largest city, are economic hubs of culture, education, and innovation.

The potash reserves in Saskatchewan are located within the Middle Devonian Prairie Evaporite Formation (a thick sequence of evaporite minerals such as halite and potash deposited in a restricted marine basin approximately 393 to 383 million years ago), in the Elk Point Basin, a region with a layered sedimentary history beginning from the Cambrian period (approximately 541 to 485 million years ago, characterized by a succession of carbonate, shale, and sandstone formations reflective of varied marine and coastal environments) (12).

As of 2023, This region is home to ten active mines and several major development projects, collectively contributing to Canada’s position as the largest producer and exporter of potash globally, as shown in the map below. In 2022, these mines produced approximately 24.6 million tonnes of potassium chloride (KCl), reflecting a steady increase in production over previous years (14).

Following are the active potash mines in Saskatchewan.

Nutrien Ltd.

1. Allan Mine: Operating since the 1960s, this mine uses conventional underground mining techniques.

2. Cory Mine: Combines conventional underground mining with solution mining for deeper deposits.

3. Lanigan Mine: Relies exclusively on conventional underground mining methods.

4. Patience Lake Mine: Transitioned to solution mining after flooding halted underground operations in the late 1980s.

5. Rocanville Mine: Near the Manitoba border, Rocanville is one of Nutrien's largest and most productive operations, utilizing underground mining techniques.

6. Vanscoy Mine: Located southwest of Saskatoon, Vanscoy has been a cornerstone of Nutrien’s operations for decades, employing underground mining methods.

The Mosaic Company

7. Belle Plaine Mine: Among the largest solution mining facilities in the world, Belle Plaine is a flagship operation.

8. Colonsay Mine: Restarted in 2022 to meet rising global demand, this mine east of Saskatoon uses underground mining.

9. Esterhazy Mine: Mosaic’s largest operation, the Esterhazy complex includes K1, K2, and the advanced K3 shafts, all employing underground mining.

K+S Potash Canada

10. Bethune Mine: Opened in 2017 near Moose Jaw, Bethune is Canada’s first new greenfield potash mine in nearly 50 years. It employs solution mining and stands as a flagship project for K+S.

BHP

11. Jansen Mine: Set to become the eleventh potash mine in Saskatchewan, the Jansen project will utilize underground mining techniques.

Canada’s Potash Transportation System

Jointly owned by Nutrien Ltd. and The Mosaic Company, Canpotex represents the unified efforts of two of Saskatchewan’s largest potash producers for Canadian potash to reach over 40 countries, supporting global agriculture and food security.

The heart of Canpotex’s transportation system lies in its extensive fleet of over 8,000 custom-designed railcars. These cars, owned by Canpotex but operated on tracks provided by Canadian Pacific (CP) and Canadian National (CN), efficiently transport potash from Saskatchewan’s mines to major port terminals in Vancouver, British Columbia; Portland, Oregon; and Saint John, New Brunswick.

This integration of specialized railcars with Canada’s robust rail infrastructure allows Canpotex to load approximately 240 vessels annually, ensuring seamless delivery to international markets. Over the years, Canpotex has invested more than $1 billion USD to enhance its logistics network, including expanding its railcar fleet and upgrading terminal facilities (15).

Capex Requirement of Potash Megaprojects

Despite the apparent simplicity of the two main processing methodologies for potash—conventional underground mining and solution mining—these projects only economically viable through economies of scale.

Economies of scale are critical because potash prices, while occasionally experiencing spikes, tend to stabilize within predictable ranges, requiring high production volumes and cost leadership to offset initial expenditures.

Solution Mining Projects

Once mineral reserves are well established, building a solution mining operation involves a series of critical steps.

First is the development of the mine, which includes establishing the wellfield and creating caverns. A wellfield is a network of wells used to inject water into potash-rich deposits, dissolving the mineral to form underground caverns filled with brine, which is then extracted and processed to recover potash. Injection and production wells are drilled strategically to ensure efficient water and brine flow, enabling the dissolution and extraction of potash-rich brine.

Next is the construction of the processing plant, where the extracted brine is treated to separate and crystallize the potash. Alongside this, waste salt management systems are implemented to handle by-products responsibly, ensuring minimal environmental impact. These steps are supported by essential infrastructure, including pipelines, power systems, and transportation networks, which are critical for sustaining efficient operations and product delivery.

A Preliminary Economic Assessment for a 3.5 Mtpa (million tonnes per annum) solution mining project, published in 2017, estimated the total cost at CA$4.03 billion. The table below presents the rounded values and normalized percentages for the various components of the project. As anticipated, the normalized costs highlight the processing plant as the primary cost driver.

Conventional Underground Mining Projects

Developing a conventional underground potash mining project involves a series of carefully planned steps.

First is the development of the mine itself, which includes sinking shafts to access the potash deposit. These shafts serve as critical access points for workers, equipment, and material transportation, and are supported by ventilation systems to ensure a safe working environment. Mining is conducted using the room-and-pillar method, where large underground rooms are excavated, leaving pillars of potash to support the overlying strata. This method ensures stability while allowing efficient extraction of the resource.

The other major part is the construction of the processing plant, where the mined potash ore undergoes a series of critical steps to transform it into marketable products. This includes raw ore handling, storage, and crushing to prepare the material for processing. The process mill building is divided into wet and dry areas. The wet area comprises attrition scrubbing, desliming, flotation, and debrining, which are essential for separating the potash from impurities. The dry area involves drying, screening, compaction, and glazing to refine the product further. These operations are complemented by waste management systems, such as tailings storage, and infrastructure for transport, power, and water supply, ensuring efficient and sustainable production.

The most recent underground potash mine in Canada with a reported capital estimate is Jansen Stage 1. The conventional underground project, designed for 4.35 Mtpa production, was estimated to cost CA$5.513 billion in 2022. The table below presents the values reported by BHP along with normalized percentages.

Notably, the underground mine features one service shaft and one production shaft, both fully lined to ensure stability and operational efficiency. The processing plant includes facilities for processing and milling, encompassing both wet and dry operations as described earlier, along with tailings processing systems. Non-process infrastructure comprises essential components such as a tailings management area, administration buildings, warehousing, workshops, utilities, and rail connections. The normalized cost distribution highlights mining and processing as two equally significant components in the economics of underground potash project.

Potash and Climate Change Adaptation: A Curious Relationship

Potash serves as a major pathway for climate adaptation in the agriculture industry, particularly in enhancing drought resistance in crops (17, 18).

At the same time, Saskatchewan experiences some of the widest swings in both the temperature and precipitation in Canada (19), posing dire risks to Canadian potash industry.

This causal dependence should highlight the importance of climate change adaptation of the potash industry for its principle stakeholders, way above its resource wealth.

As a key source of potassium, potash plays a critical role in regulating water uptake and retention in plants, enabling them to maintain hydration even under arid conditions. These benefits make potash indispensable for farming in regions prone to erratic rainfall patterns, where the impacts of climate change are intensifying.

Furthermore, potassium aids in detoxifying reactive oxygen species generated under water stress, safeguarding photosynthesis and ensuring continued crop productivity. This dual role of conserving water and sustaining metabolic functions underscores potassium's pivotal role in developing climate-resilient agriculture systems, making potash an indispensable resource for global food security in a warming world.

The potash mining industry in Saskatchewan faces significant risks from climate change that directly challenge its operations and sustainability.

One of the most pressing issues is the increasing frequency of excessive moisture events and flooding. Such occurrences result in higher volumes of brine for disposal, escalating operational costs and creating logistical challenges for storage and treatment. This amplifies the necessity for robust infrastructure investments, including enhanced flood bypass systems and deep-injection wells, to mitigate operational disruptions and environmental risks.

Water scarcity and prolonged droughts is another major climate risk. Water is integral to potash extraction and processing, with both conventional and solution mining methods relying heavily on its availability. Climate-induced reductions in water quantity and quality could compromise operations and strain existing water supply systems. Mining companies have begun implementing water reuse and recycling initiatives, optimizing their consumption patterns, and exploring innovative sourcing options such as utilizing treated wastewater (20).

Bracing against long-term climate projections possibilities require a need for integrated adaptation strategies, regulatory support, industry collaboration, and scientific innovation to sustain the sector amid changing climatic conditions.

This underscores the broader challenge of aligning an old yet fast growing industry.

Question then Becomes

Our road trip through Canada’s potash heartland was a rewarding journey, offering firsthand insight into an industry that sustains global agriculture.

We saw first hand, the mines, the infrastructure networks, and landscapes that together deepened our appreciation for Canada’s vital role in feeding the world. It also gave us interesting questions to think about.

Canada’s potash industry, despite its economic importance, often lingers in the shadow of resource giants like oil and energy. This relative obscurity diminishes its potential significance as a major driver of development, both directly within Canada and indirectly across the world, as well as its role in a climate change adaptation pathway for the agriculture industry.

How can the Potash industry better present its role in global food security and secure its rightful place alongside Canada’s other flagship resource sectors? What steps, from strategic marketing to stronger policy support, could elevate its visibility and impact?

Potash offers an important opportunity for the mining industry to reshape its reputation by highlighting its contributions to alleviating hunger and promoting sustainable agriculture. How can the Canadian mining industry better embrace this narrative?

Notes

Refer to this article as below:

• Zangeneh, P., Hassanpour R., F. (2025), “Canada’s Potash Legacy: A Road Trip Through 40% of the World’s Food Fertilizer Production,” EPM Research Letters.

Join the EPM Network to access insights, influence our research, and connect with a community shaping the industry's future.

Please support us by sharing this article with your friends and colleagues.

Subscribe now

References

1. Al Rawashdeh, R. World peak potash: An analytical study. Resource. Policy 69, 101834 (2020).

2. Al Rawashdeh, R. & Maxwell, P. Analysing the world potash industry. Resource. Policy 41, 143–151 (2014).

3. Word Bank Group. https://datacatalog.worldbank.org/.

4. Broughton, P. L. Economic geology of southern Saskatchewan potash mines. Ore Geol. Rev. 113, 103117 (2019).

5. Fertilizer Canada - Growing Stronger. Fertilizer Canada, https://fertilizercanada.ca/.

6. Potash. https://www.thecanadianencyclopedia.ca/en/article/potash.

7. CMM. A Great Future in Potash. Canadian Mining Magazine https://canadianminingmagazine.com/a-great-future-in-potash/ (2024).

8. Saskatchewan Potash History – by John Burton (The Encyclopedia of Saskatchewan). Republic of Mining https://republicofmining.com/2014/03/18/saskatchewan-potash-history-by-john-burton-the-encyclopedia-of Saskatchewan-unknown-date/ (2014).

9. Saskatchewan Research Council. https://www.src.sk.ca/blog/helping-industry-produce-premium-potash-support-global-agricultural-community.

10. Saskatchewan potash nourishing the earth. http://saskmining.ca/ckfinder/userfiles/files/SMA-Potash-Infographic-2022-Stats-WEB.pdf

11. Saskatchewan potash vital for world food, https://www.resourceworks.com/saskatchewan-potash-vital.

12. Scientific Investigations Report. https://portergeo.com.au/database/mineinfo.asp?mineid=mn1801 (2023).

13. Government of Saskatchewan, https://geohub.saskatchewan.ca/datasets/saskatchewan::mine-locations/

14. Canada, N. R. Potash facts. https://natural-resources.canada.ca/minerals-mining/mining-data-statistics-and-analysis/minerals-metals-facts/potash-facts/20521 (2018).

15. Canpotex. https://www.canpotex.com/news/canpotex-invests-its-world-class-supply-chain-acquiring-1300-railcars-national-steel-car.

16. BHP, https://www.bhp.com/what-we-do/global-locations/canada/jansen

17. https://www.icl-group.com/blog/the-role-of-potash-in-sustainable-agriculture/

18. Sardans, J. and Peñuelas, J. (2015), Potassium stoichiometry and global change. Global Ecology and Biogeography, 24: 261-275. https://doi.org/10.1111/geb.12259

19. Haave, D. (2012). WUQWATR Climate Extremes Preparedness Workshop Final Report. Wascana Upper Qu’Appelle Watershed Association Taking Responsibility (WUQWATR)

20. Pittman, J., Ford, J. & Pearce, T., 2014. Adaptation to climate change and potash mining in Saskatchewan, Canadian Electronic Library. Ottawa, Ontario. Retrieved from https://coilink.org/20.500.12592/x10208 on 15 Jan 2025. COI: 20.500.12592/x10208.