AI
Yelp's AI Assistant Blends Native Controls with Server-Driven UI for Faster Updates
Yelp adapted its server-driven UI framework to power conversational AI while keeping navigation and input natively built, enabling quicker feature rollouts across iOS and Android.
InfoQ
Yelp's AI Assistant uses a pragmatic hybrid approach to user interface design, pairing native mobile components with server-controlled conversation rendering. The strategy lets the platform deploy feature changes to users instantly without requiring app updates—a key advantage when iterating on AI-driven experiences.
The Hybrid Architecture#
Rather than go fully server-driven or fully native, Yelp split the difference. The company extended its CHAOS framework (Content Hosting Architecture with Optimization Strategies)—a multi-platform UI infrastructure—to handle conversation content, while leaving navigation, message input, and state management as native code. This separation acknowledges a practical tradeoff: server-driven UIs excel at rapid iteration, but native controls provide better performance and familiarity for core interactions.
As Daniel Andrade Groppe, Services Tech Lead at Yelp, noted, extending CHAOS to support the text input fields and network requests the Assistant needed would have slowed launch. A hybrid implementation kept the project on track while preserving the benefits of both approaches.
How Events Power the Conversation#
The system decouples view configuration from message data. When a chat session begins, the client fetches the CHAOS layout once, then updates it with a stream of events—each representing a new message, a typing indicator, a quick-reply button, or a system action. This event-driven design means the backend can inject new conversation states without the client refetching the entire UI blueprint. The backend can even send actions to disable input or trigger navigation once a task completes.
The client handles display logic, input state, and error recovery natively, reducing latency for user interactions while still letting the server drive what users see and do in the conversation area.
Pattern in the Industry#
Airbnb pioneered a similar model with its Ghost Platform, which decouples presentation from client code. Both approaches enable teams to roll out UI changes instantly to all users—web, iOS, and Android—without shipping new app versions, a major advantage for rapidly evolving AI features.
Tradeoffs and Future#
Yelp acknowledges the approach introduced complexity: engineers needed to learn both the Assistant system and CHAOS, and shared components can still diverge across platforms. But the architecture also unlocked reuse; the Android foundation powered a pilot product called , demonstrating that the investment paid dividends. The company says this foundation now underpins a next-generation Assistant that weaves business information with actionable user workflows.

