All roles
for product managers

AI for product managers

PMs spend most days shaping context — turning messy customer signals into shipping priorities, turning founder vibes into specs, turning data into decisions. That's exactly what AI is good at: structuring chaos into something a team can act on. The trap is letting AI do the prioritizing for you. The lever is letting it surface options faster so you can decide better.

8 use cases·7 tools·30-min starter

What AI handles well

Customer interview synthesis

The problem. You ran 10 user interviews. The patterns are in there but extracting them takes a day of re-listening.

What AI does. Paste transcripts into AI. Ask for: recurring pain words, current workarounds, moments of strong emotion, contradictions across interviews. Patterns surface in 30 minutes.

Use the promptTools:Granola or Otter for recording. Claude for synthesis.

PRD first draft from messy thinking

The problem. You have rough notes: customer feedback, eng concerns, designer sketches, business asks. Writing a PRD that ties them together takes a half-day.

What AI does. Feed the messy notes into the strategy doc prompt. Get a structured first draft (problem, proposal, why this approach, success criteria, risks, open questions). Edit the parts only you can write.

Decision frameworks for hard calls

The problem. Build vs buy. Kill or invest in this feature. Path A vs Path B. You're going in circles.

What AI does. Use the decision framework prompt. It surfaces what you're really deciding, options you're missing, reversibility, and the smallest test that earns you information cheaply.

Pre-mortem before committing

The problem. You're about to launch a major initiative. Six months in, you find out about the failure mode you should have caught.

What AI does. Run a pre-mortem before you commit. Imagine the project failed — what happened? AI helps you list specific failure modes, early warning signs, and what to kill if it doesn't work in 4 weeks.

Build a metric tree for your goal

The problem. Leadership wants you to "improve activation." You don't know what to measure or where the leverage is.

What AI does. Use the metric tree prompt. Break the goal into mathematical drivers, leading indicators, and the 1-2 metrics worth obsessing over given your stage.

Hypothesis tests for "is this real?"

The problem. Customers say they'd pay for X. You're about to commit a quarter to building it. You're not sure if "would you pay" survey responses are signal.

What AI does. Use the hypothesis-test prompt. It surfaces the right test (LOI? prototype? pricing experiment?), confounds to control for, and the cheaper smell-test you can run first.

Internal announcements (deprecation, change, restructure)

The problem. You're killing a feature, changing a workflow, or restructuring the team. You need to announce it without confusion.

What AI does. Use the internal announcement prompt. TL;DR up top, what's changing, why (honestly), what people need to do, where to ask questions.

Competitive analysis without the bloat

The problem. Leadership wants competitive analysis. You don't have time for a 30-page deck. You also don't want to do it on intuition.

What AI does. Use the compare-sources prompt. Surface what competitors agree on (boring), where they disagree (interesting), and what they're NOT saying that you could own.

Your AI stack

Start with the foundation. Add specialized tools as the work calls for them.

Foundation LLM

Claude
Best for synthesis (interview transcripts, doc analysis), nuanced writing (PRDs, strategy docs), and decision frameworks. The PM's default LLM in 2026.
ChatGPT
Strong for structured tasks, list generation, quick drafts. Custom GPTs let you bake in your team context.

Specialized add-ons

Granola or Otter
AI meeting notes for customer calls. Frees you from typing during the call so you can actually listen.
Notion AI
Drafts and summaries inside the docs your team already uses. Lower friction than tool-switching.
Linear
Issue tracker with built-in AI for prioritization, summary generation, status updates.
Productboard or Canny
Customer feedback aggregation with AI tagging and prioritization.
Maze or UserTesting
For unmoderated user research. AI transcripts + sentiment analysis at scale.

Prompts ready to use

Get started in 30 minutes

1

Set up AI meeting notes for your next 5 customer calls

10 min

Granola, Otter, or whatever your team approves. The goal: you stop typing in calls and can actually pay attention to the human in front of you.

2

Build a Claude Project for your product context

10 min

Load: your ICP, your product description, key competitor positioning, your team's priorities. Now any prompt is pre-loaded with context.

3

Run the pre-mortem prompt on your current biggest initiative

10 min

Before you commit further, surface the failure modes. You'll either uncover something worth changing or confirm your plan — both are useful.

Common mistakes

  • AI-summarizing customer calls and never re-listening to a single one. The summary kills nuance — the moment of frustration, the unexpected detail, the contradiction. Listen to at least one call per week unsummarized.

  • Using AI to generate prioritization. AI doesn't know your strategic context or political constraints. It'll give you a "balanced" priority list. Reality requires opinionated tradeoffs only you can make.

  • AI-generated PRDs that read like every other AI-generated PRD. Engineers and designers can spot it. The first paragraph is where YOUR judgment shows up.

  • Letting "synthesis" replace listening. AI compresses. Compression loses information. Use AI to surface patterns, then go back to source material for the decisions that matter.

  • Pasting customer data (names, companies, conversations) into public AI tools. Use enterprise tiers or anonymize first.

Related roles