Lab · exploring
A critique partner that argues back
Can an AI critique agent give a lean design team the rigor of a senior review — without the senior being in the room?
July 10, 2026
Hypothesis
A well-scoped critique agent, grounded in our own design principles, raises the floor of feedback for a small team more than another design tool would.
Outcome
Early — the agent is good at surfacing consistency and accessibility gaps, weak on strategic trade-offs. Documenting where the human review still has to lead.
This is a sample entry. Replace it with your first real experiment — keep the hypothesis / outcome framing, it’s what makes the lab read as rigor, not a demo reel.
The question
Small teams don’t have a senior reviewer in every critique. I wanted to test whether an AI agent, grounded in a specific set of design principles, could raise the floor of feedback — catching the obvious before a human spends attention on the hard part.
What I built
A critique agent scoped to one product surface, fed our principles and a component inventory, prompted to review flows against them.
What I learned
- Strong on the mechanical: spacing drift, contrast, inconsistent labels.
- Weak on the strategic: it can’t yet weigh a trade-off it wasn’t told about.
- The real value wasn’t replacement — it was giving juniors a faster first pass so the human review could start higher up the ladder.
The interesting design-leadership question isn’t “can AI critique.” It’s where in the critique ritual you place it so the team gets sharper instead of dependent.
← Back to the lab