What It Takes to Make AI Booking Work at Scale
Discover how HotelPlanner built Reservations.ai from an internal tool into a voice-driven travel agent, boosting conversion rates from 2% to 16%.
Stock photo for illustration only, not from the actual event
- HotelPlanner built Reservations.ai internally before opening it to outside partners
- The system trained on 10 million recorded calls gathered during the pandemic
- Conversion rates climbed from 1-2% to human-level performance of 14-16%
- Roughly 30% of customers still request to speak with a human agent
The travel industry is largely still testing AI bookings in controlled environments. Companies that can prove the technology works across live bookings at scale could gain a valuable head start in the market.
Deloitte’s 2026 State of AI in the Enterprise report found that only 25% of organizations had moved 40% of their AI pilots into production, while 37% were still using AI at a surface level without materially changing underlying processes.
Stock photo for illustration only, not from the actual event
HotelPlanner came to market from a different starting point by developing Reservations.ai, a voice-driven, fully integrated booking system designed to handle its own multi-supplier marketplace before offering it as a standalone product to outside partners.
Skift Studio spoke with Tim Hentschel, co-CEO of HotelPlanner, to explore what it takes to make AI booking technology work at scale, and what travel companies should consider when deciding whether to build or buy.
"Reservations.ai started during the travel industry’s post-pandemic recovery."
Tim Hentschel
Training conversational AI on millions of actual customer service calls provides a distinct advantage in mastering accents, colloquialisms, and complex travel itineraries. Integrating voice technology directly with multi-supplier inventory represents a major evolution for online travel agencies.
Stock photo for illustration only, not from the actual event
Initial results were challenging with conversion rates around 1–2%, but the team raised performance to a human-level rate of roughly 14–16% over the course of a year by refining back-end technology to better understand real customer speech patterns.
Source: Skift
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment