I design cloud products people can actually operate.

I inherited three designers covering fourteen product teams, which meant everybody was a service desk and nobody owned an outcome. The fix was not more headcount first — it was deciding what a designer here is accountable for, then hiring against that.

I wrote the loop we still use: portfolio screen, a paid craft exercise on a real (declassified) problem, a working session with the squad they would join, and a leadership conversation about how they handle disagreement. No whiteboard trivia.

A system nobody adopts is a hobby. We shipped only the twelve components that appeared in every product, then let demand pull the rest — and published a deprecation calendar so teams could plan instead of being surprised.

Governance is a biweekly council, not a queue. Anyone can contribute; the council decides what becomes canonical and what gets deprecated, in public, with the reasoning written down.

Quality does not survive on taste alone; it survives on a repeatable room. Critique here is weekly, timeboxed, and structured so the most senior person speaks last — otherwise the loudest opinion becomes the decision.

The presenter states the decision they need. Feedback that does not serve that decision gets written down for later instead of derailing the session.

Most design failures here were not craft failures — they were sequencing failures. Design arrived after the commitment was made, so the only available moves were cosmetic.

I moved design into the shaping phase: a designer sits in the problem definition and leaves with the same understanding of constraints as the engineers. In exchange, we commit to a scope we can actually finish.