Among the announcements from the "What's new in Flutter" session at I/O 2026, the most interesting for anyone designing conversational experiences is the release of Flutter GenUI: an open-source SDK that lets AI agents compose interfaces at runtime, selecting and parameterizing widgets from the app's existing catalog rather than returning plain text.
How it works
The model does not generate code or pixels: it emits JSON mapped to a portion of Flutter's widget tree, and the framework renders it instantly. When the user interacts with the generated UI, state changes are fed back to the agent, creating a high-bandwidth loop in which every gesture becomes input for the next conversational turn. This is the difference between an agent that explains options and an agent that puts them in front of you.
Why it matters
The promise is twofold. For consumer products it means interfaces that adapt to user intent: a booking request generates a booking tile, a product-comparison question generates a comparative card, without each screen being pre-built. For enterprise products it removes the design bottleneck for fast-changing flows. On 9to5Google and in the official Google post, GenUI is positioned as a key step in the shift from AI-assisted to AI-agentic application development.
Availability
GenUI is open source and officially documented at docs.flutter.dev/ai/genui. It supports mobile, desktop, and web, consistent with Flutter's multi-platform nature. The community is already producing integration examples with platforms like FlutterFlow to build end-to-end agentic flows.