AI won't replace product managers — it changes them

Every few months someone asks, half-joking, whether AI is going to make product managers obsolete. I understand the instinct — a lot of what fills a PM’s calendar looks automatable from the outside: writing requirements, summarizing customer calls, drafting the roadmap slide, triaging a backlog.

But most of that was never the actual job. It was the paperwork around the job. The job was deciding what the requirements should say in the first place — which trade-off matters more this quarter, which customer signal is real versus loud, which “quick win” quietly makes the platform worse in six months. AI is genuinely good at the paperwork. It is not making those calls for you, and I don’t think it’s going to.

What’s actually changing is where a PM’s time goes. I’ve spent the last year building small AI-assisted tools — a valuation workbench that turns judgment calls into structured, defensible reasoning; a categorization model that proposes where a transaction belongs so a person doesn’t have to sort it by hand. In both cases, the AI’s job is to compress the mechanical part of a decision so a person can spend their attention on the part that actually requires judgment. That’s the pattern I’d bet on generally: not AI replacing the decision-maker, but AI clearing enough of the noise that the decision-maker can actually see the decision.

The PMs who struggle with this shift, I think, will be the ones whose value was mostly in producing artifacts — the doc, the deck, the ticket. The PMs who do well will be the ones whose value was always in judgment, because that part doesn’t compress. If anything, it gets more visible once the paperwork stops hiding it.