A masterclass on designing products where AI answers, ranks, or acts, presented at FITC Web Unleashed, Toronto, on 28 October 2026. The session argues that the design problem has moved beyond prompts and chat interfaces to a question of ownership: who owns an AI's answer once someone acts on it, and who fixes it when it is wrong. It works through four documented public incidents — Moffatt v. Air Canada (2024), Amazon's experimental recruiting engine (2014–17), Knight Capital (2012), and the Replit agent incident (2025) — and derives four frameworks from them: Map (the Interaction Stack / CLEAR model), Set Authority (the Autonomy Spectrum), Design Trust, and Gate the Release (the AI Experience Audit). Each framework is applied back to a case, measured with a named metric, and demonstrated on a worked example: a meeting assistant taken through all four tools. The deck is shared with the full speaker script in the notes, as presented. The companion handout, The Pattern Deck: A Field Guide for Teams Shipping AI, collects the seventeen design patterns and the Figma kit that builds them.
No takes yet. Share an insight, caveat, or question.
Mathew Graham (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: