Summary
At the AI Engineer conference, Gus Iwanaga, General Manager at commercetools, argues that modern software UIs have accumulated decades of cognitive debt — layers of features, navigation flows, and onboarding requirements that still burden users even as AI enables faster shipping. His talk makes the case that generative, intent-driven interfaces can replace static screen paradigms entirely, allowing software to adapt to the user rather than the other way around.
Iwanaga walks through a live demo of commercetools’ generative UX system, showing how AI agents dynamically assemble UI components based on user intent rather than fixed navigation paths. He outlines three distinct UI rendering protocols — from fully opinionated “control” components to more fluid AI-composed layouts — and explains why B2B SaaS products serving highly configurable workflows require more flexibility than consumer apps like Booking.com. The system lets an agent select from a component catalog and render context-appropriate interfaces on the fly, with human approval before changes go live.
The talk closes with hard-won engineering lessons: maintaining quality in non-deterministic UI generation at scale, managing configuration complexity, and aligning stakeholders on what “AI-guided UX” actually means in practice. Iwanaga also references emerging UI protocol standards — whose founders were present at the same conference — as key infrastructure for the next generation of adaptive enterprise software.
📺 Source: AI Engineer · Published September 01, 2026
🏷️ Format: Keynote Launch
