There's a version of the AI-in-product-management conversation that goes: AI will automate the low-value work, freeing PMs to do more strategic thinking. This is partially true and mostly beside the point. The more important observation is that AI removes friction from output — and for a PM who was already doing great work, that's powerful. For a PM whose thinking was fuzzy to begin with, it just produces fuzzy thinking at greater volume.
What a bad PM with AI looks like
More PRDs. Written faster, formatted well, full of plausible-sounding user stories that aren't connected to any real customer insight. More roadmap proposals, generated quickly, prioritized by criteria that sound strategic but were assembled from prompts rather than hard thinking. More competitive analyses, more market research summaries, more slide decks.
All of it confident. Most of it misaligned. The bad PM who used to take a week to produce work that was directionally wrong can now produce it in an afternoon — and ship it to the engineering team before anyone has had a chance to ask whether it's right.
Speed without direction is just a faster way to end up somewhere you didn't want to be.
What a great PM with AI looks like
They use AI to accelerate the thinking they were already doing well. They use it to pressure-test — "here's my argument, what am I missing?" They use it to synthesize customer feedback faster so they can spend more time in actual customer conversations. They use it to get to a draft quickly so they can spend their energy editing and sharpening rather than generating.
The great PM isn't asking AI to do their thinking. They're asking it to do the parts of the work that require processing power rather than judgment — and preserving their energy for the parts that require judgment, which is where the actual value lives.
"The skill isn't knowing how to use the tools. It's knowing what question to ask. That's judgment. AI doesn't give you judgment."
What actually needs to be developed
Judgment. Customer empathy. The ability to walk out of ten customer conversations with a synthesized point of view about what matters and what doesn't — and the confidence to act on it even when it's unpopular. Strategic clarity about where you're going and why, clear enough that it can absorb uncertainty without falling apart.
None of these come from AI tools. They come from doing the work, being wrong, paying attention to why, and building up enough reps that the pattern recognition starts to work reliably. They come from time spent with real customers, not with user personas. They come from being in the room when hard decisions get made, not from reading about how hard decisions get made.
The organizations that will use AI well in product development are the ones that invest in developing real product judgment in their people — and then let those people use AI as a force multiplier. The ones that don't will find that AI gave their existing gaps a megaphone.
