“The skills I cultivated for 15 years are now available for free.” David Pereira wrote that this week, and it landed harder than most PM hot takes do. Not because it’s provocative. Because it’s accurate: research, synthesis, structured thinking, competitive analysis, documentation. Everything you spent years learning to do well is now something any PM can generate in minutes.
The question that follows is the one most PM content is too nervous to answer: what’s left?
Not are PMs safe — that’s the wrong question. The right question is: which parts of the role are now table stakes, and which parts were always the actual job?
What AI can do now
The commoditised parts are real. AI can research a market faster than any PM. It can draft a PRD, a brief, a competitor analysis, a set of user stories. It can synthesise customer feedback and structure a roadmap.
Everything in that list was once a skill. It is now infrastructure.
A PM who spent years building value around those practices has a real problem. Not because they will be replaced — but because the floor has risen. Everyone has access to what used to be advanced.
What it didn’t replace
Taste. Knowing what’s worth building is not a research problem. It’s a judgment call made with incomplete information, shaped by pattern recognition that takes years to develop. AI can tell you what users asked for. It cannot tell you what they’d love if they saw it. The PM who knows the difference isn’t the one who runs the best synthesis prompt — it’s the one who has been wrong enough times to know how wrong feels.
Context. AI has no memory of how the last decision was made, what was tried two years ago, why that feature was killed, what the real reason was that the roadmap shifted in Q3. The PM who carries organisational context — who knows the history, the dynamics, the things that aren’t in any document — is not replaceable by something that starts fresh every session.
Relationships. Trust is built over time, not generated. The engineer who tells you something is going wrong before it blows up does it because you’ve earned that conversation. The stakeholder who gives you room to move does it because they believe in your judgment. AI can draft the message. It cannot build the relationship the message relies on.
The courage to kill. Saying no to a feature that looked good on paper. Cancelling a project halfway through. Telling a VP their pet idea isn’t worth building. AI optimises, it doesn’t disappoint people. The PM who can do that — and carry the room through the discomfort — is doing something no model can be trained to replicate.
The uncomfortable implication
Some PMs are nervous about AI because they have always derived their value from the parts that are now commoditised. The wrong response is to double down on process like more thorough PRDs, tighter documentation, better-structured reviews. That is optimising for the thing AI already does better.
The right response is to ask honestly: was I ever really doing judgment work? And if not, how do I start?
The PMs who aren’t nervous are the ones whose value was always in the harder-to-name things. AI hasn’t threatened them — it’s removed the noise that obscured what they were actually good at. The Real PM Test I wrote about last month is essentially this question applied to hiring. The same logic applies to careers.
What this means for how we hire
If the commoditised skills are no longer differentiating, hiring criteria need to change. A PM who writes crisp PRDs is not impressive in 2026. A PM who makes hard calls, carries context across a long project, and builds trust with an engineering team: that is what you are now actually paying for.
Most job descriptions haven’t caught up. Most interviews still test the wrong things. That is going to cost organisations the people they most need, to the ones who already have.
AI has not replaced product management. It has clarified what product management actually is.
Leave a Reply