An AI project manager needs two layers of competence: the project leadership PMI’s PMP certifies, and the ability to carry out the four functions of NIST’s AI Risk Management Framework (Govern, Map, Measure and Manage) on every AI system they deliver. NIST says what an organization should do to manage AI risk; below, each function is translated into what the person running the project must answer for, with the interview evidence to ask for.
Every item traces to a public document, so you can check it against the source and fit it to your own job architecture.
| Key facts
8% → 26% Business Environment’s share of the PMP exam after PMI’s July 2026 update, which also added AI to project-based scenarios. Source: PMI. 4 functions Govern, Map, Measure and Manage, the core of NIST’s voluntary AI Risk Management Framework (AI RMF 1.0, January 2023), which NIST is now revising. Source: NIST. 21 tasks O*NET’s task list for Information Technology Project Managers (15-1299.09), the O*NET occupation closest to this role. Source: O*NET OnLine. 80%+ Share of AI projects that fail by some estimates; the most common root cause RAND found was misunderstanding the problem. Source: RAND. |
Why map the role to public standards
Your risk, audit and PMO teams already work from these documents. A role defined in their terms is easier to approve, staff and audit, and a candidate’s experience is easier to check against it.
The standards also leave a gap this role fills. The NIST AI RMF and ISO/IEC 42001 describe what an organization should do and call for clear roles and responsibilities, but neither names the job title that does the work on a given project. Assigning the project-level work to one accountable person is how many organizations close that gap.
The five areas at a glance
| Area | Public source | What the AI project manager answers for |
| Project leadership | PMI PMP exam outline; O*NET 15-1299.09 | Scope, schedule, cost, risk, stakeholders, vendors and team leadership on every AI initiative. |
| Govern | NIST AI RMF: GOVERN 2, GOVERN 6 | Who approves what, what the system may do unattended, the escalation path, the owner after go-live, and third-party terms. |
| Map | NIST AI RMF: MAP 1, MAP 3 | The problem, intended use, users, expected benefits and costs, and the success measure, agreed before build. |
| Measure | NIST AI RMF: MEASURE 1–3 | Evaluation criteria and acceptance tests before launch, and how performance and risks are tracked over time. |
| Manage | NIST AI RMF: MANAGE 1, 2.4, 3, 4.1; BCG 10-20-70 | Go/no-go decisions, third-party risk, monitoring, incident response, change management, and scale, fix or retire. |
Project leadership (the floor)
What the standard says. O*NET lists 21 tasks for Information Technology Project Managers, from managing execution against budget, schedule and scope to selecting vendors and running risk assessments. PMI’s July 2026 PMP update keeps the exam’s core principles and raises Business Environment from 8% to 26%.
What the AI project manager answers for:
- Scope, schedule, budget and risk for cross-functional AI initiatives with outside vendors.
- Vendor selection and contract scope, with procurement.
- Status, risk and value reporting to sponsors in plain language.
Interview evidence: “Walk me through a project where scope changed mid-stream. What changed in the plan, and who approved it?”
Govern: assign authority and accountability
What the standard says. NIST calls GOVERN “a cross-cutting function that is infused throughout AI risk management.” GOVERN 2 asks for accountability structures so the appropriate teams and individuals are empowered and responsible, and GOVERN 6 asks for policies covering third-party software and data. ISO/IEC 42001 likewise calls for defined roles and responsibilities. Neither names the job title that does this on each project.
What the AI project manager answers for:
- What each system may do without human review, when it must escalate to a person, and who that person is, set with risk and compliance partners.
- A named internal owner for every system before the vendor or project team rolls off.
- Third-party terms (data use, support, handoff) built into the plan rather than discovered after launch.
Interview evidence: “For an AI system you delivered, what was it allowed to do without a person reviewing it, and who decided?”
Map: frame the use case
What the standard says. NIST’s MAP function “establishes the context to frame risks related to an AI system.” Under MAP 3, the system’s capabilities, targeted usage, goals and expected benefits and costs should be understood. RAND’s research points the same way: the most common root cause of AI project failure it found was a misunderstanding of the problem the project was meant to solve.
What the AI project manager answers for:
- The problem, the decision the system supports, and the measurable result that counts as success, agreed with business owners before build.
- The intended use, the users and the limits of each system, written down.
- The business case, revisited at each stage gate, including a recommendation to stop when a use case can’t meet it.
Interview evidence: “Tell me about an AI or automation idea you recommended not doing. What told you?”
Measure: decide how it’s judged
What the standard says. NIST’s MEASURE function uses “quantitative, qualitative, or mixed-method tools” to assess AI risk. Its categories ask that appropriate methods and metrics are identified and applied (MEASURE 1), that systems are evaluated (MEASURE 2), and that “mechanisms for tracking identified AI risks over time are in place” (MEASURE 3).
What the AI project manager answers for:
- Evaluation criteria and acceptance tests agreed before launch, with the people who will sign off.
- Monitoring in place at launch, not added after the first problem.
- Performance and risk tracked over time against the success measures set under Map.
Interview evidence: “How did you decide an AI system was good enough to launch, and who agreed?”
Manage: launch, run and retire
What the standard says. NIST’s MANAGE function “entails allocating risk resources to mapped and measured risks on a regular basis.” It covers mechanisms to “supersede, disengage, or deactivate” systems that don’t perform as intended (MANAGE 2.4), third-party risks (MANAGE 3), and post-deployment monitoring that includes incident response, recovery and change management (MANAGE 4.1). BCG’s 10-20-70 approach explains why the people side matters: it attributes 70% of AI value to people and processes.
What the AI project manager answers for:
- The go or no-go decision, made on the evidence from Measure.
- Incident response and a switch-off path, planned before launch.
- Change management and training for the teams whose work the system changes.
- The handoff from vendor engineers to the internal owner, and the call to scale, fix or retire each system based on results.
Interview evidence: “Pick an AI system six months after launch. What were you measuring, and what did you change because of it?”
Govern and Manage meet at the handoff, often the week a vendor’s engineers roll off, when the learning leaves with them unless an internal owner is already in place. We cover that moment in who owns your AI after go-live.
How to use it
- Ask one evidence question per area and listen for a real system, a real decision and who made it. Our AI project manager job description template groups its responsibilities the same way.
- Leveling and pay. We suggest treating ownership across vendors as the seniority signal: a role that owns all four NIST areas for a portfolio is senior; one that supports them on a single project is earlier-career. The AI project manager salary guide shows what each costs.
- Score your current project managers against all five areas and plan around the gaps. We’d expect gaps in the four NIST areas even among strong PMs, because standard PM roles haven’t asked for them.
- What not to do. Don’t turn it into a checklist of tools or model types. The areas describe judgment, and judgment shows up in evidence from real projects.
| Score your bench
For each project manager on your team, which of the five areas could they show evidence for from a real AI project in the last 12 months? Any area where nobody can is the one your next AI deployment is most likely to stall on. |
What the PMP covers, and what it doesn’t yet
The first area is the PMP’s territory. In our reading, the 2026 update also reaches toward Map and Manage through its heavier Business Environment weighting. PMI’s update adds AI to exam scenarios and keeps the core principles the same; it doesn’t set out a separate skill set for deploying AI systems. That’s why the PMP is the right floor for an AI project manager, with NIST’s four functions built on top of it.
If you’d rather build the role than hire it
Some employers will grow AI project managers from their own staff rather than compete for experienced hires.
Master of Project Academy sponsors the nation’s first registered apprenticeship dedicated to AI Project Management, registered by the Maryland Apprenticeship and Training Council (MATC #2575, September 8, 2026) under the registered occupation Information Technology Project Manager. The registration was announced via Business Wire.
The AI Project Management apprenticeship explains how the program works and how employers take part.
Frequently asked questions
What competencies does an AI project manager need?
Two layers. First, the project leadership the PMP certifies: scope, schedule, cost, risk, stakeholders and vendors. Second, the four functions of NIST’s AI Risk Management Framework applied to each system: Govern (assign authority and accountability), Map (frame the use case), Measure (decide how it’s judged) and Manage (launch, run and retire it).
Does an AI project manager need to know how to code?
Usually not. The role needs enough technical literacy to understand how AI systems are evaluated and monitored, and to challenge engineers and vendors on those points. Building models is a different job.
Is the PMP still relevant for AI project managers?
Yes. It’s the floor for the role. PMI’s July 2026 update raised Business Environment from 8% to 26% of the exam and added AI to project-based scenarios. The AI-specific work, as NIST’s framework describes it, is built on top of it.
Which frameworks apply to AI projects?
Two common reference points are the NIST AI Risk Management Framework, a voluntary U.S. framework organized around Govern, Map, Measure and Manage, and ISO/IEC 42001, an international management-system standard for AI. Sector regulations apply on top. An AI project manager applies them on each project.
| Can’t hire an AI project manager? Build one.
The AI Project Management apprenticeship, registered under the Information Technology Project Manager occupation (MATC #2575), develops one of your own employees into the role, on your projects, as a full-time employee from day one. |
P.S. Looking for an AI project manager role yourself? Our guide to how to become an AI project manager covers both ways in.
About Master of Project Academy
Founded in 2012, Master of Project Academy has served 500,000+ enrollments across 180+ countries in project management training. It sponsors the AI Project Management apprenticeship, registered under the Information Technology Project Manager occupation (MATC #2575). Learn more at masterofproject.com/p/aipmapprenticeship.
Sources
- NIST — AI Risk Management Framework (AI RMF 1.0, NIST AI 100-1)
- NIST — AI Risk Management Framework overview and revision status
- PMI — The new PMP exam (July 2026)
- O*NET OnLine — 15-1299.09 Information Technology Project Managers
- ISO — ISO/IEC 42001:2023 AI management systems
- RAND — The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed (2024)
- BCG — What’s Driving AI Value (Jan 26, 2026)