When it comes to the potential for innovation and disruption that artificial intelligence brings, the travel sector appears to be poised for one of the industry's most profound transformations since the invention of the commercial airline.
Traditional online travel booking relies on rigid page structures, drop-down menus and filters. AI-driven travel booking isn't about replacing humans; it's about removing friction. It's about blending planning, booking and inspiration into a single conversation.
Kayak.ai is designed to emulate natural human conversation and the kind of planning you might do with a knowledgeable agent or friend. By moving from static search interfaces to dynamic, agent-driven experiences, the travel experience now allows users to inquire about everything from hotels to car rentals, destination dining options, and activity suggestions -- all within a single conversational thread.
"Right now, it's more like an attended booking within the chat," he said. "You'll still see what you're booking, confirm the price, and maybe enter your credit card. But over time, this could get more and more semi-automated. Maybe you leave your details, and the booking happens in the background."
Behind the scenes, Kayak's existing integrations with major airlines, hotel chains and global distribution systems (GDS) give it a competitive edge, he said.
"We already have very great direct airline integrations... and we're connected to GDS where all travel availability lives," Keller said. "That gives us a lot of already well-integrated content to play with."
While consumer travel grabs headlines, there may be even greater potential for agentic AI in the corporate sector. Enterprise travel has high frequency, strict compliance needs and a demand for instant service, so it is a prime testing ground for AI that plans, books and reacts in real time.
"These are travelers who fly every Monday and return every Friday," Keller said. "They want to be productive. They do high-volume bookings and want to spend the least time doing it. But they also need the flexibility to cancel, rebook and adjust plans on the fly."
Enterprise use also offers a sandbox for testing more advanced features with a degree of built-in tolerance.
"They have more opportunities to cancel content, to change content," Keller said. "So, it's a great test bed to try agent features with some safeguards."
Keller said he believes users who are more comfortable with natural language interfaces than with traditional websites are setting the standard for the future of digital travel. He said he also sees agentic AI as a tool to democratize complex travel planning.
"In the future, I think the most successful travel companies will be the ones that turn AI into not just a back-end tool but a front-end experience," Keller said.
Still, with automation comes risk, especially when personal data, payment information and loyalty credentials are involved. Kayak, long known for its transactional depth, sees this as a manageable challenge, he said.
"We have had our credit card processing environment for probably 10 years," Keller said. "So, a lot of this is in our DNA. We feel good about having great building blocks."
A user profile that includes stored payment information and loyalty accounts will be central to enabling agentic booking in the future.
"That's where it really becomes important that the agent knows about you and can put in your card when it comes time to book," he said.
The next frontier is individualized, responsive trip planning, Keller said.
"Travel is about planning and reasoning," he said. "If I just plan a trip to Berlin, I need to pick the flight before the hotel. Otherwise, I may book a hotel for dates when I'm not even there."
That's why, in Keller's vision of the future, users won't just query the system. They'll delegate complex planning entirely, he said.
"Maybe I say, 'I want to spend three weeks in Europe,' and the agent says, 'For $3,000, everything's taken care of -- click buy,'" Keller said. "It's a reasonable itinerary and everything works out. That's agents talking to each other, our own personalization, and real planning."
