Writing

A maturity model for how design teams actually adopt AI — from mechanical assistance to autonomous systems — and why the jump out of Level 1 is organizational, not technical.

We designed agents as features. They're colleagues. Job descriptions, onboarding, permissions, a manager, a probation period in staging — what three years of shipping agent products taught us to unlearn.

My research on how voice design shapes trust in AI advisors — begun as a 2016 thesis, now in review at the International Journal of Social Robotics (with C. Fausett & D. Nathan-Roberts). Medium pitch and fast speech measurably raise trust; people defer to machines more on functional decisions than social ones; and people conform to confident, explained recommendations regardless of accuracy — which should worry all of us building explainable AI.

AI lets anyone ship a prototype — which is exactly the problem. Why prototypes collapse the solution space while PRDs keep it open, and the collaboration table my teams use to keep both honest.