The AI Product Manager's Playbook
Where AI gives a product manager real leverage today (research, insights, data, prototyping) and where to keep human judgment (vision and the strategy bet), plus the one prompting tip that matters most.
Guillaume Rufenacht
AI Product Manager · Lisbon
AI can supercharge a product manager across strategy, design, and execution, which is exactly why it feels overwhelming: it can touch everything, so where do you actually start? The useful answer isn’t “use AI more.” It’s knowing the handful of places it gives you real leverage today, and the two places you should still trust your own judgment.
I’m an AI product manager by trade, so this is both a synthesis of the best framework I’ve seen and how I actually work. Here’s the map.
Key takeaways
- The PM role has four dimensions: vision, strategy, design, execution. AI helps unevenly across them.
- AI makes you faster, smarter (it critiques your thinking), and gives you superpowers (analyst and researcher work).
- Lean in for research, customer insights, data analysis, prototyping, and execution hygiene.
- Hold the line on vision and the actual strategy bets, AI gives consensus answers, and non-consensus bets win.
- The best prompting tip: shape the output with an exemplar of what 'great' looks like.
The frame: four dimensions of the PM role
A PM owns four things: the vision (how the world is better if you succeed), the strategy (how you win your market), the design (customer and data insights through to specs and prototypes), and the execution (actually shipping). AI helps enormously in some of these and barely in others, and knowing which is the whole skill.
Where to lean in
AI doesn’t just save time, it makes you smarter and hands you capabilities outside your role. The high-leverage areas:
Strategy research and critique. AI is excellent at market and competitor research, and even better at playing devil’s advocate against a strategy you wrote. Tell it not to be nice and it will find the holes, the unfocused audience, the moat that isn’t a moat, faster than most review meetings.
Customer insights and data. Synthesize survey verbatims, run AI-moderated interviews at a scale you couldn’t staff, turn a CSV into a segmented analysis with significance testing, answer data questions in plain English instead of queuing them for a data team. This is the “superpower” layer: you do the analyst’s and researcher’s work yourself.
Prototyping and execution hygiene. Build interactive prototypes to test before engineering is involved, and let AI handle meeting summaries and agendas. None of this is glamorous; all of it compounds.
Where to hold the line
AI is weakest exactly where the role is most valuable: vision and the actual strategy bet. It returns consensus-driven, pattern-matched answers, and the decisions that win markets are usually non-consensus, the performance EV when everyone else optimized for cheap commuting. Use AI to research and pressure-test, but the bold, opinionated call stays human. The quant in me likes this framing: AI gives you the base rate; your job is to find where the base rate is wrong.
The one prompting tip that matters most
The PM becomes the bottleneck (in a good way)
When analysis that took a week takes minutes, the constraint moves from “can I get this done” to “what should I ask.” You run NPS continuously instead of quarterly; you run a hundred interviews instead of ten. The leverage is real, but it accrues to the PM with curiosity, taste, and end-to-end ownership, which is the same bet I make in everything I build. It’s also why these roles are merging, more on that in how Anthropic’s team works.
The takeaway
This is how I work as an AI PM, taking products from 0 to 1 across strategy, design, and build. See my experience or get in touch.
Frequently asked questions
Where should PMs use AI today?
Research and critique for strategy, customer insights and data analysis and prototyping for design, and the hygiene of execution (summaries, agendas). Keep vision and the actual strategy bets for human judgment.
Why isn't AI good at product vision or strategy?
Because it returns consensus-driven, pattern-matched answers, and the bets that win a market are usually non-consensus. Use AI to research and pressure-test, not to decide the bold direction.
What's the single best prompting tip for PMs?
Shape the output with an exemplar. Give the model a great reference, a respected book or your own best work, and ask it to apply those standards, instead of asking cold for a generic answer.
Work with me
Want a system like this built for your pipeline?
I help teams take AI from a clever prototype to dependable production, outbound engines, lead intelligence, and the LLM pipelines underneath. See what I have shipped or get in touch.