There is no AI domain and no sustainability domain. Both are woven into existing tasks, so nobody will ask you to define a neural network or recite an ESG framework. You will be dropped into a project where an AI tool or a sustainability requirement is part of the situation and asked what a good project manager does.
Two topics arrived in scope with the July 2026 exam content outline, and both are being prepared for badly — usually by candidates trying to learn a subject that is not being tested.
There is no AI domain and no sustainability domain. Both are trends the job task analysis validated as part of modern project work, and the exam weaves them into existing tasks rather than testing them as separate subjects.
That changes what preparation looks like completely. Nobody will ask you to define a neural network or grade a company's ESG report. They will drop you into a project where an AI tool or a sustainability requirement is part of the situation and ask what a good project manager does — which is the same question the rest of the exam asks.
Artificial intelligence appears as a tool the project manager uses, not a force that runs the project. The questions test judgment with AI, and they cluster around three ideas.
Tooling. AI can help plan, estimate, analyse status and surface information faster than a person can. But it informs decisions; it does not own them. The project manager who lets an AI tool make the call has handed away accountability that is theirs to keep.
Data. An AI output is only as good as the data behind it. A confident forecast built on incomplete or stale project data is a confident wrong answer, and checking the inputs is part of using the tool.
Ethics and governance. AI outputs can carry bias, can be hard to explain, and can leak confidential data into places it should not go. Using AI responsibly means addressing the bias, being able to explain an AI-influenced decision to stakeholders, and protecting the data you feed in.
The trap here is the AI oracle: treating an output as a verdict to obey rather than an input to weigh. The tool is fast and sounds certain, so the tempting answer follows it without checking the data, questioning the recommendation, or owning the decision. If the AI is wrong, it is still the project manager's call that failed.
Sustainability, like AI, has no domain of its own. It is a consideration threaded through the project's decisions, and on the exam it wears several familiar hats. Recognising which hat it is wearing tells you which task's logic applies.
It is a critical information requirement when the project must track sustainability targets, not just cost and schedule. It is part of value when a deliverable's worth to its customers now includes how sustainable it is. It is a compliance requirement when a regulation mandates a standard. It is a risk when an environmental or ESG threat could stall the work. It is a cost-of-quality question when you decide whether to build sustainability in from the start or bolt it on at the end. And it is an external trend when the market shifts toward sustainable options.
None of that is a separate skill. It is the tasks you already know, applied to a requirement that happens to be about sustainability.
The trap here is greenwashing: claiming or reporting sustainability the project has not actually delivered. The label goes on the marketing, the report shows the target met, the box is ticked — while the practice underneath does not match the claim. A hollow claim is worse than none: it collapses on inspection, damages trust, and often breaks a compliance rule.
Neither gets questions labelled AI or sustainability. Both ride inside tasks you have already studied, which is why hunting for them as topics does not work.
| Where it appears | AI | Sustainability |
|---|---|---|
| Critical information requirements | Data-informed decisions in planning | Tracking sustainability targets alongside cost and schedule |
| Value-based delivery | — | Worth to the customer that now includes how sustainable it is |
| Quality | — | Cost of quality: build it in, or bolt it on at the end |
| Project status and metrics | Reviewed metrics and artefacts | — |
| Communication and governance | Transparent reporting of AI-influenced decisions | — |
| Compliance | Confidentiality of the data you feed in | A regulation mandating a sustainability standard |
| Risk | Bias, explainability, data leakage | An environmental or ESG threat that could stall the work |
| External business environment | Technology shifts that reshape scope or backlog | A market shift toward sustainable options |
In every case, the tested skill is the task decision, and it takes the same four shapes.
Verify the evidence. Whether the input is an AI forecast or a sustainability claim, the first question is whether it rests on something real. An unverified output and an unverified claim fail the same way.
Protect the requirement. A sustainability requirement is a requirement. It is planned, built in, measured and delivered like any other, and treating it as optional or deferrable is the mistake.
Govern the action. Someone must own the decision, be able to explain it, and be accountable for it. That is true of a decision informed by an AI tool and of a decision to report a sustainability target as met.
Measure the result. Real sustainability is delivered and measured, not asserted; a decision taken with AI assistance is still judged by its outcome.
If you can do those four things, you can answer these questions without knowing anything about machine learning or ESG reporting frameworks — which is exactly the point.
Do not study AI. Do not study sustainability. Study the tasks, and then practise questions that place an AI tool or a sustainability requirement inside them.
The useful drill is to take a scenario you can already answer and add one of these elements to it. A schedule recovery question becomes an AI question when the recommendation comes from a scheduling assistant; the correct answer does not change much, but the temptation does. A procurement question becomes a sustainability question when one criterion is environmental; the logic of procurement still applies.
Watch for the emotional pull, because both topics generate it. AI outputs sound authoritative. Sustainability claims sound virtuous. Both push you toward accepting something you have not verified, which is the same failure wearing two costumes.
And treat any material that gives you an AI syllabus or an ESG framework to memorise with suspicion. That is not what is in scope, and time spent there is time not spent on the tasks that carry the marks. Our own guide handles both as situational questions inside the existing tasks — ten set in AI-assisted environments, ten in sustainability-driven decisions — rather than as subjects with their own theory.
As always, PMI's current published outline governs what is in scope.
No. There is no AI domain and no sustainability domain. Both are woven into the existing tasks of the exam content outline rather than tested as separate subjects.
No. You will not be asked to define a neural network. You will be asked what a project manager should do when an AI tool is part of the situation — verify the inputs, weigh the recommendation, and own the decision.
Treating an AI output as a verdict to obey rather than an input to weigh. The tool is fast and sounds certain, so the tempting answer follows it without checking the data or owning the decision. Accountability never transferred.
No. The exam does not ask you to grade a company's ESG report. It asks whether you treat a sustainability requirement with the same seriousness as any other — planned, built in, measured and delivered.
Claiming or reporting sustainability the project has not delivered. It is tested because it is tempting — it captures the market benefit without the work — and because a hollow claim collapses on inspection and often breaks a compliance rule.
PMI publishes weights at the domain level only — People 33 percent, Process 41 percent, Business Environment 26 percent — and none for individual topics. We will not invent a figure. What is certain is that both are in scope and both appear inside existing tasks.
Rules change. Where a figure or a procedure can move, the issuing agency’s current published instructions win over anything here.
