Every search, every booking, every price update on Booking.com flows through a single federated graph - the context layer between their platform and every human, app, and travel agent that needs to reason over it. And now it needs to work for AI agents too.
Christian Ernst has been building that context layer for seven years. He helped it grow into a company-wide platform - and watched the governance model break when teams went their own way. Now he's rebuilding governance with the help of LLMs, discovering that hallucination is actually a feature when you need realistic mock data at scale, and exploring how to open Booking's inventory to AI agents.
In this episode, Christian takes us inside one of the world's largest API platforms - and to the payoff that matters most: the graph built for Booking's web and mobile apps turned out to be the graph AI agents needed too. Same types. Same verbs. Built for humans, ready for agents.
What you'll hear
Why agents need a context layer, not raw API endpoints
How Booking turns LLM hallucination into auto-generated API mocks
Why comms - not code - is the biggest lever in any modernization program
Meet your speakers
