Before Figma Make and Claude, a fully clickable concept took two days. Now it takes an afternoon. The instinct is to ship faster. The better move is to widen the search.
I now run three to five parallel directions through the first review instead of one polished candidate. Stakeholders argue against real artefacts instead of slide bullets, and the cheapest direction to abandon is the one you only spent an hour on.
What did not change: the questions. What user, what job, what evidence. AI is a faster pencil, not a sharper one.