Selfbook

Selfbook is a next-generation hotel commerce platform that powers direct booking for hospitality brands worldwide. Their Direct Distribution Network reaches approximately 500 million users.

Backed by American Express, Tiger Global, and Affirm, Selfbook delivers a 43% revenue lift and 39% conversion increase for hotel partners. The next frontier: replacing the fragmented search-and-book ritual with a single intelligent conversation.

Booking a hotel still looks like it did a decade ago. Users bounce between search engines, comparison sites, and payment screens — each handoff bleeding intent and conversion.

The technical challenge was steep: real-time integration with hotel inventory, dynamic UI generation, persistent memory across sessions, and location-aware search — all at the scale of hundreds of millions of users.

We built an agentic AI travel assistant on LangGraph with LangSmith for observability. The system dynamically generates UI widgets based on conversational context: maps for exploration, ranked lists for comparison, room cards for narrowing, and payment flows for booking.

AWS Lambdas handle compute, durable chat memory preserves context across sessions, and deep integrations with Travelport and Google Maps APIs connect the agent to live inventory and geographic intelligence.

The AI travel agent went from initial concept to functional MVP in one month. Durable memory, real-time inventory integration, adaptive UI generation, and location-aware intelligence are all live.

For Selfbook’s hotel partners, the agent represents a fundamentally different guest acquisition channel — hotels surface naturally through conversation, matched to traveller intent rather than bid price.