For a decade, product teams were organized around a handoff: designers make mockups, engineers implement them. AI-native development is dissolving that boundary from both sides. Engineers ship UI without opening a design tool; designers ship working prototypes without waiting for sprint capacity; agents generate plausible interfaces from a sentence. The handoff is dying — but accountability for quality, consistency, accessibility, and user trust hasn’t gone anywhere.

This firechat puts a Principal Product Designer and a senior engineer on stage for a candid, unscripted conversation about what this shift actually looks like inside a product company shipping to thousands of customers:

The 70% problem. AI produces a convincing first draft of anything — UI, copy, code. Who is responsible for the last 30%, and how do teams stop “plausible” from shipping as “done”?
Design systems as APIs. When agents generate interfaces, a design system stops being a component library for humans and becomes a contract that machines consume: tokens, patterns, copy rules, accessibility standards. What does it take to make design knowledge machine-readable — and what silently breaks when you don’t?
Encoding judgment. Real production examples of converting institutional design knowledge into AI-consumable skills — including how a design team automated the rebranding of 400+ product illustrations across a legacy codebase with an AI pipeline (extraction → classification → ticket generation → automated application), and what that taught us about where humans stay in the loop.
What we each stopped doing. Which parts of the designer’s and the engineer’s job have already been absorbed by AI — and which parts became more valuable.
Where we disagree. Speed vs. craft, autonomy vs. review, and who should be allowed to ship UI in 2027.