The PM who ignores AI won't be replaced by AI — they'll be replaced by a PM who used it to get four times more done in the same day. That's the entire premise behind AI-PM: not automating judgment, but offloading the repetitive, time-consuming parts of the job so there's more time left for the parts that actually need a human — negotiation, prioritisation, and reading a room.
Below are five prompts I use and teach in AI-PM sessions. They're deliberately structured, not one-liners — the difference between a mediocre AI output and a genuinely useful one is almost always in how much context and structure you give the prompt, not which model you use.
1. The Status Report Generator
Turn raw updates — sprint notes, a messy list of what happened this week, a few Slack threads — into a clean, structured stakeholder update in seconds.
"Act as a project manager preparing a weekly status update for [audience — e.g. steering committee]. Using the raw notes below, produce a status report with these sections: Overall Status (Green/Amber/Red with one-line justification), Key Accomplishments This Week, Risks & Issues (with owner and mitigation), Upcoming Milestones, and Decisions Needed From Leadership. Keep it under 300 words, plain business English, no jargon. Raw notes: [paste notes]"
2. The Risk Identification Sweep
Before a risk register review, run your project description through this prompt to surface risks a busy team might not think to log themselves.
"You are an experienced risk manager reviewing a project before a risk workshop. Based on this project summary [paste scope, timeline, team structure, key dependencies], identify 10 potential risks across these categories: schedule, scope, resource/people, technical, vendor/third-party, and stakeholder. For each, give a one-line description, likely impact (High/Medium/Low), and a suggested early-warning signal to watch for."
3. The Stakeholder Translation Layer
The same update needs to sound different for an executive sponsor than for the delivery team. This prompt does that translation instantly.
"Rewrite the following project update for a [C-level executive / delivery team / client] audience. Priorities for this audience: [e.g. business impact and risk exposure only, no technical detail]. Keep it to 150 words, lead with the headline outcome, and end with exactly one clear ask if there is one. Original update: [paste]"
4. The NFR & Requirements Drafting Assistant
Non-functional requirements are the ones teams forget until they cause a production incident. This prompt forces a structured first pass.
"Act as a business analyst drafting non-functional requirements for [brief system/feature description]. Cover these categories at minimum: performance, security, scalability, availability, usability, and compliance. For each NFR, write it in the format 'The system shall [requirement], measured by [metric], under [condition].' Flag any category where you need more information from me rather than guessing."
5. The Retrospective Synthesiser
Turn a pile of raw retro sticky notes or a transcript into themes and action items instead of a wall of unstructured feedback.
"Below is raw output from a sprint/project retrospective (sticky notes or transcript). Group the input into themes (e.g. communication, process, tooling, people). For each theme, summarise the sentiment in one line and propose one concrete, owner-assignable action item. Flag any comment that appears more than twice as a priority pattern. Raw input: [paste]"
Never let AI output go to a stakeholder unreviewed. These prompts are accelerators for the first draft — the judgment on what to keep, cut, or soften before it reaches a sponsor is still the PM's job. That review step is exactly where the "human in the loop" earns their seat at the table.
Building your own arsenal
These five cover the recurring, time-consuming parts of the job. The real skill is learning to structure a prompt the way you'd brief a sharp junior analyst — give them the audience, the format, the constraints, and the raw material, and they'll generally get you 80% of the way there.