Travel Industry AI Agents Expose Data Gaps in Hotel Booking Systems

Expedia Group’s first AI booking chatbots, including the retired Romie, failed because the assistant’s answers were not tied to the company’s live inventory, rates and reservations. Chief product officer Shilpa Ranganathan said the bot “gave generic, ungrounded responses,” leading travelers to abandon the conversation and eroding trust. The fix required rebuilding the data plumbing and evaluation framework so the chatbot could answer from real inventory, a process that took about a year and involved little work on the model itself. Ranganathan likened the problem to trying to build a tower on “a foundation of sand.”

A new report from hotel‑technology firm ZentrumHub, based on more than 1.5 million real bookings, finds that AI agents are beginning to book hotel rooms on behalf of travelers. The report estimates that by the end of 2027, 5‑8 % of OTA bookings could be made by AI agents rather than humans. As co‑founder Abhinav Sinha explains, “An AI agent doesn’t scroll through pretty pictures. It checks prices and availability in milliseconds and books from whoever answers fastest with the cleanest data.” The study warns that hotels with slow or messy inventory risk being bypassed by these agents, underscoring the industry‑wide need for accurate, real‑time data.