Speaking as someone who uses AI every working day, including on personal projects, my honest read is that AI has made the product management role more important, but it has changed what the role actually is, and it hasn't changed it evenly across levels.
I spend two to three hours a day in an AI assistant for user and product research, competitor and data analysis, PRD and test-case documentation, and funnel analysis across multiple products. What used to take hours now takes minutes. What used to take a full day now takes a few hours. I also prototype directly now, instead of writing a discovery document and showing a rough mock to stakeholders, I build something clickable and have the conversation around that instead.
What's actually left after the tools do their part
Once the production layer, research, drafting, first-pass prototypes, is handled, what's left is deciding what to build and what not to build. Reading the funnel and recognising that a drop-off isn't a UX problem but a segmentation problem. Getting Risk, Ops, and Engineering aligned. Bringing a growth team that wants something different around the same evidence. Taking the trade-off call when every side has a legitimate claim.
AI cannot make those calls. They come down to product sense built over years, and they arrive faster now than they used to, which means the judgement behind them gets scrutinised more, not less.
The role isn't shrinking. The floor is rising, and the floor is rising faster than the title is.
Where I think the optimistic framing misses something
A common counterpoint, and a fair one, is that this isn't happening evenly. AI has made day-to-day PM work faster and easier across the board, but instead of that uniformly making PMs more valuable, a lot of organisations are using the same leverage to compress the ladder. Junior PMs are increasingly expected to carry mid-level scope with a skip manager guiding from above, on a junior salary. The middle layer thins out while the production work gets automated out from under it.
Both things are true at once. The role's ceiling is rising, because the judgement layer is genuinely more valuable when it's unblocked by less production drag. And for a meaningful slice of the workforce, the floor is being pushed up onto people who aren't compensated for the jump. The company captures a lot of the value AI created before the individual PM does.
What I'd actually tell a PM at any level
If most of your value used to come from writing documents, running analysis, and producing first drafts, that value is compressing fast, and it doesn't matter how good you are at it. If your value comes from reading ambiguous signals correctly, building trust across functions that don't naturally agree, and being accountable for a call when the evidence is incomplete, that value just got a lot more visible, because the noise around it disappeared.
The honest move isn't to get faster at the things AI already does well. It's to get better at the one thing it still can't do, deciding, under real disagreement, what's actually worth building next.