Booking's CEO says handling your guest just needs "a little more compute"
Booking Holdings has cut what it costs to answer a guest's problem by double digits twice in six months. There is no place to rate the machine that answers, so the only review in the transaction is the one with your name on it.
by Markus Busch
What he says it takes to answer a guest. Booking Holdings chief executive Glenn Fogel told the Wall Street Journal on August 14 how he intends to end the hold music. "I want it so nobody ever has to wait on line ever. How will that happen? Well, when you're using AI agents, you don't need to hire more people. You just need a little more compute." Fewer customer contacts per booking, he added, and "our actual costs for customer service per booking are going down."
Then he described what the machine is answering. Asked how he weighs the AI spending against the customer experience, Fogel went to the complaints that reach his own inbox. "Most of the problems are issues where we're not the supplier of the flight. We are not the hotel that has a noisy air conditioner. We are not the rental car that has a flat tire." And he says where they land: "we have to deal with the airline, the hotel, the car rental company."
Read it again with your property in it. The volume Booking.com is automating is complaints about your rooms.
The number has moved twice in six months. On the February 18 call, chief financial officer Ewout Steenbergen put customer service cost per booking down ten percent against the year before, while bookings rose ten percent. "Gen AI is everywhere except for in the P&L," he said. "I'm very proud that we can point to a line item in the P&L where we actually do have meaningful results." On the August 4 call the same cost fell at a double-digit rate again, this time on voice automation.
The other claim he made has nothing behind it. Told that readers write in frustrated at having to talk to an AI agent before they can reach anybody, Fogel said the research backs him: "We actually have data that shows people prefer the AI agent versus having to wait for a human." Booking Holdings has not published it. Two claims about the same machine came out of that interview. One is audited, dated and reported to investors. The other is a sentence.
Nobody rates the machine. Here is the part that makes it yours. A guest spends her evening going around with a bot about a room that was too warm. When it's over, where does that go? Look for the place to rate Booking.com's service desk. There isn't one. No star on the voice agent, no score on the chat. The whole transaction contains exactly one review form, it reaches her after checkout, and it has your name at the top of it.
She won't separate the two, and nobody would. The room was too warm, getting help was a maze, and all of it happened under the heading of a trip to your hotel.
What it costs you isn't measured. Nobody has put a number on how far a bad service-desk experience drags a property score, and this piece isn't going to invent one. The structure is what's documented: the friction is being automated on a schedule reported to investors every quarter, and the only scoreboard in the transaction is yours.
Next time a guest mentions she already tried Booking.com about the room, ask her how that went. Then read your reviews for that week.
Booking.com saved the call. You'll meet her at breakfast.
Read also: The real cost of an OTA booking isn't the commission — it's the guest
Enjoying this analysis? Hospitality.today delivers daily insights on hotel distribution, AI trends, and travel commerce — straight to your inbox. Subscribe for free at Hospitality.today →