06/13/2026
AI is not a growth lever. It's a mirror.
I say that because every time a founder asks for an 'AI feature' they mean: please make my users act like they already want this.
I've seen teams bolt on models to avoid hard product decisions. Replace pricing clarity with personalization. Replace strategy with shiny automation.
You get better demos, not better retention.
I believe useful AI requires constraints. Not more capabilities. Constraints force you to decide: who exactly benefits, what exactly changes, what cost you accept.
I've sat through workshops where the output was a dozen user journeys generated by a model and zero decisions about trade-offs.
People treat hallucinations like bugs. They are signals. A model making stuff up often reveals unclear definitions in your data or in your brief.
So my working rule: add AI where it raises the cost of doing the wrong thing. Raise friction when it matters.
Practically, that looks like small features that expose trade-offs to users, strict success metrics, and ruthless pruning of edges.
Not because AI is hard. Because humans are lazy about decisions. And that's the real product risk.