AI Won't Replace Product Managers. But It Will Expose Weak Ones.

"AI will replace Product Managers because it can write the PRD and user stories."

I keep hearing some version of this. It is framed as an obvious logical step: if the main output of a PM is documentation, and AI can produce documentation, then the PM becomes redundant.

The argument has a surface plausibility that makes it worth taking seriously. And then dismantling.

The document is not the value

A PRD is a record of thinking. The thinking is the work.

The thinking requires someone to have spoken to enough users to understand what they actually need versus what they say they need. It requires someone to have sat in enough stakeholder conversations to know which requirements are real constraints and which are preferences dressed up as mandates. It requires someone to have made a call about what not to build, and to have held that decision when the pressure came.

None of that is in the document. The document is what you write after you've done it.

An AI model can pattern-match a PRD from a brief. It can write user stories from a spec. It is genuinely useful for that. But the brief still has to come from somewhere, and the spec still has to be right, and the call about what the spec should say is still a human judgement.

What the model can't be in the room for

The value of a PM compounds over exposure. It comes from being in the room when a stakeholder says one thing and means another. From the user interview where the question you planned to ask turned out to be the wrong one. From the product review where the metric looked fine and something still felt off.

That pattern recognition develops through contact with reality, over time, across many decisions. It doesn't transfer through a brief.

A model can be given the output of that experience — the notes, the research, the decision log — and produce something plausible from it. But it cannot have the experience itself. It cannot feel the discomfort that tells you the question was wrong. It cannot notice the thing that wasn't said.

Why weak product thinking will show up faster

Here is what AI will actually do to product management: it will make the gap between good and bad PM thinking much more visible, much more quickly.

A strong PM will use the tools to move faster through the parts that were always mechanical. The research synthesis, the story writing, the documentation. They'll spend more time on the parts that actually require judgement, and they'll get to decisions faster.

A weak PM will use the tools to produce more documentation more quickly. The PRDs will look better. The stories will be more numerous and more consistently formatted. And none of it will be grounded in the understanding that makes those documents useful.

The output will look similar. The underlying thinking won't be.

In organisations that evaluate product management by the quality of the artefacts, this is hard to detect. In organisations that evaluate it by outcomes — what shipped, what worked, what changed — it shows up fast.

What actually gets replaced

AI does remove real work from product management. The administrative overhead, the template-filling, the first-draft generation that used to eat significant time. That work going away is unambiguously good.

What it doesn't remove is the need for someone to listen properly, understand what is really going on, and then decide.

If anything, as the documentation becomes easier to produce, the listening and deciding become more of the job, not less. The question is whether the people doing product management are ready for a role that is less about producing things and more about knowing what to produce and why.

That's what it always should have been.