POST 03 · PRODUCT SENSE

Product Management: The Judgment and Product Sense That Makes the Difference

BY SHYAM GOLE · 6 MIN READ

AI made generating things cheap. Ten feature ideas, five roadmap options, three pricing models — thirty seconds each. Which means the scarce thing in product management is no longer producing options. It's choosing between them. That choice — made with incomplete data, real stakes, and no undo button — is judgment, and it's the one part of the job AI hasn't touched.

What judgment actually looked like for me

• Scoping Quinn to three jobs, not thirty. AI could brainstorm fifty things an assistant might do. The interviews said owners were stuck on three: content, images, scheduling. Cutting the other forty-seven — including some genuinely good ideas — is what made month-one adoption possible.
• Charging from day one. Every playbook says launch free and convert later. The call was that a paying customer's complaint teaches more than a free user's compliment — and the first 16 paying customers proved the demand was real.
• Killing a feature that demoed well. The hardest calls aren't bad ideas — those kill themselves. They're good ideas that don't fit the strategy, the team, or the moment. "Good idea, not now" is a complete sentence.

Where product sense comes from

Nobody's born with it, and no course installs it. Mine came from three places: reps with real users (every interview deposits a little pattern-recognition you can't get second-hand), shipping and being wrong (the features that flopped taught me more than the ones that worked), and owning outcomes (when the number is your number, you stop admiring options and start weighing them).

Why AI can't make the call

Not because it isn't smart. Because judgment needs things a model doesn't have: the context of your specific users, team, and history; accountability for the outcome; and taste — a view of what should exist, not just what's statistically likely. AI predicts the average answer. Product bets that pay off are, almost by definition, not the average answer. So I use AI to widen the funnel and sharpen the evidence — and I keep the filter human.