How should design critique change when AI makes the first draft?

Answered by Falk Gottlob · Reviewed · 2 min read

THE SHORT ANSWER

Split it and plan it as capacity. Giuseppe Soda and colleagues tracked more than 12,000 workers through a Microsoft Copilot rollout and, as Jakob Nielsen reports it, adopters added about 1.4 hours of collaboration a month, much of it trusted coworkers checking AI output. That's critique with no format. Give it one: the AI generates ten options and an AI critic runs the first pass against the rubric, then people re-rank the severity and kill options with one written line of why. The kill reasons go into the file the design agents read.

Split it, and put it on the capacity plan. When the first draft is the AI's, the scarce work is judging, and judging is what a critique is for.

The evidence. Giuseppe Soda, Alessandro Iorio, and Leonardo Rizzo tracked more than 12,000 knowledge workers for 13 months through a Microsoft Copilot rollout. As Jakob Nielsen reports it, adopters started about 1.4 more hours of collaboration a month and took AI output to coworkers they trusted, a circle he calls the vetwork. That's critique with no format and no record.

The split. The AI generates ten options that differ on purpose, and an AI critic runs the first pass against the rubric. People re-rank the severity and kill options with one written line of why. Don't put a room of people on line edits of one generated screen: Michelle Vaccaro's meta-analysis of 420 studies found that arrangement loses to the AI alone.

The log. The kill reasons go into the file the design agents read, so the same mistake isn't critiqued twice.

The full argument is part 7 of The Design Transition, the series runs in order on the hub, and The Product OS shows why one critique practice across product, design, product marketing, and the field sits under the CPO.

SOURCES

THE LONG VERSION

RELATED ANSWERS