AI can generate interface and flow at speed, which means production is no longer where the value sits.
Judgement is. Which problem is worth solving, whether the assumption beneath a feature actually holds, and what a system should never do.
I build Product Design functions around that shift.
Design moves from shipping fixed screens to shipping constraint systems: the rules, safety rails and evaluation criteria that govern what AI is allowed to generate. Research stops being a deliverable and becomes the evidence loop that keeps judgement honest.
Three principles hold that judgement in place.
i.What's worth making
The scarcest call in product work is deciding which problem deserves to exist, before anyone designs a thing. AI makes building cheap. It does not make this judgement for you.
ii.Refusal is a feature
The design isn't only what a system generates. It's what it declines to. Knowing what a product should never do, and building the rails that hold that line, is the work, not a guardrail bolted on at the end.
iii.Calibration
Opinion is cheap. Judgement calibrated against continuous user evidence is not. Research is the fuel; the call is the craft.