The USD$1.6tn global travel market is about to get the greatest shakeup since the dawn of the internet. The major Online Travel Agents (OTAs) like Booking.com, Expedia and Airbnb are bracing themselves for a quantum change and it’s one they cannot control. As consumers move from booking travel through these OTAs to using ChatGPT and other AI models instead, the process of booking, the type of bookings and the commercials surrounding distribution is about to get a complete shake-up.
Taking hotels for example, it is widely accepted that in the near future most hotel bookings will be made agentically. That is to say that customers will use their AI Agent to search, discover, consider, hyper-personalize, negotiate and book their hotel stays. While many industry analysts are assuming that the OTAs will again dominate this distribution, I would like to show here why this is simply not the case.
The last two decades have seen the major OTAs dominate hotel bookings and their superpower has been marketing (they spend billions of dollars annually on Google search) and persuasion (the use of urgency, scarcity and risk mitigation). Today Booking.com (Booking Holdings) sits at a market cap which exceeds Marriott, Hilton, IHG, Hyatt and Wyndham – combined!
The competitive marketing prowess of the OTAs has, up to now, been based on old consumer behavior of search and booking. In the AI Agentic future however it is no longer a battle of marketing – it’s all about data. If hotels play this right, then direct bookings will dominate, and distribution costs will fall dramatically.
To understand how hotels can play to win we need to look at how data takes center stage in an AI driven world. With respect to hotels, there are five dimensions of data that will govern how and where agentic bookings will be made.
The Five Dimensions of Data are:
Speed. Data speed should be such as to enable agents to deliver hyper-personalized, instant experiences, turning inquiries into conversions faster than traditional OTAs.
Structure. When data is unstructured, AI tries to bridge the gaps and it infers. In doing so it often hallucinates, and this causes an array of problems.
So, data speed and structure are paramount for all future players. They are table stakes if you like and an unlikely source of competitive advantage for anyone. The next three dimensions of data however are where hotel-direct wins big in my opinion.
Depth. Hyper-personalized agentic requests require deep hotel data. Just imagine where the data for this sample request would come from:
“I need a four- star hotel in central London on Monday Dec 1st. A room on a high floor and away from the elevator, with an iron and ironing board and a pull-out bed for a child aged three. A dinner reservation for 8pm on the 1st of December and a babysitting service from 7:30 to 10:30 that evening.”
In identifying the data required to fulfill and then hyper personalize a price for the above, you quickly see how shallow OTA data is.
Trust. Trust is arguably the biggest challenge for Agentic AI. How does AI know the hotel is real, the room pictures are real, the price is not hallucinated, and the booking terms and conditions will not change? roomangel Foundation (an industry owned not-for-profit) is building the verified agentic web for hospitality. Provided free of charge to hotels, this is a simple to implement cryptographic verification system for hotels that verifies all hotel data. This ensures AI agents that the provenance and integrity of the data consumed in agentic conversations can be trusted completely.
Value. How will agentic AI know if the room price quoted represents good value for money? roomangel tags every hotel rate with a value-for-money code (a value signal) that is objective, transparent and has consumer regulatory oversight. This ensures agents that they are not overpaying or are perhaps saving a percentage from an objectively calculated, expected rate. This is a powerful tool for hotel direct and is diametrically opposed to the commission model that so heavily influences OTA search results.
Loyalty and rewards systems in this agentic world will become part of the overall value data set I believe but we will need to rethink what that agentic loyalty marketplace may look like. There is no technical reason for example that a customer cannot define the rewards that are meaningful to them. For example, “For this hotel booking I would like rewards in the form of Emirates Skywards points, Amazon vouchers or XYZ.”
In conclusion let me quote one of the great first principle thinkers on AI; Marc Andreessen. In an interview in June this year he said:
“AI is the new computer…all incumbents are going to get nuked and
everything is going to get rebuilt.”
As we move then towards the agentic AI future of distribution and everything is about to be completely reimagined and rebuilt, let’s not be deterministic and assume that the OTAs that have dominated in the past will do so again.
Instead let’s ask a radical but very real question.
“If, as an industry we deliver better, deeper, more trusted data to AI agents as I have described above, then… Why are OTAs needed at all?”
