Marty Cagan argues AI changes product work in two distinct ways: adding AI to a product, and using AI to change how the product is built2. On the second, delivery has become cheap, so the bottleneck has moved to discovering a solution worth building5.
01What changes in building
- Faster code or design generation does not automatically produce better outcomes2.
- Cagan describes prototyping at 10 to 20 per week, which was already possible before generative AI5.
- Cagan and Bob Baxley note that faster tools only help when paired with strong judgment from product managers and designers8.
02What changes in product roles
- Cagan expects most PMs will need to build AI-powered products, much as mobile PM became a baseline skill3.
- Lenny Rachitsky expects AI to hit high-level strategy and goal-setting hardest, while soft skills like influence and communication grow in value7.
- Teresa Torres says building AI features resembles orchestrating a team of interns, with decomposed tasks and oversight4.
03What PMs should watch
- Cagan urges PMs to understand AI risks, set quality thresholds, and design for trust3.
- Torres recommends logging full traces and using structured evals, since thumbs up/down signals are not enough for non-deterministic outputs4.
- PM Atlas doesn't cover the long-term effect on team size or hiring.
The short answer: AI makes building cheap, so the product manager's judgment about what to build matters most.
Written by PM Atlas from the cited notes only, drawing on Marty Cagan, Teresa Torres, Lenny Rachitsky. Quotes are short excerpts; read the originals for the full argument.