The Agents Are Here (and Almost No One Is Ready)

Adapted from my recent talk at the Skift Data and AI Summit Europe in London

Hello Research Friends,

Last week, I booked my first agentic business trip. Over text message, I worked with Instinct, an AI agent, to find trains and flights that fit my schedule, budget and loyalty preferences. I searched, compared, booked, and paid over text via agent. I'm flying it this week and will report back how it goes. But here’s the most important point: The agents are here, and almost no one is ready.

Here are the data I shared at the Skift Data and AI Summit Europe, hosted in London. I used a framework we're building on at Skift Research to describe the travel industry: the travel stack. There are four layers to it, and AI is going to reshape every one of them:

Let's take them in order.

Consumers: Trust is the gating factor

AI usage is high. Our report, Skift Global Travel Insights: What's New for 2026?, found that 62% of global travelers are now familiar with AI planning tools. But AI hasn't earned the booking… yet.

Our joint study with McKinsey & Co., Winning Hearts in an Age of Infinite Travel Choices, found that people are still the most trusted source of travel advice: friends and family first, online reviews second, and creators third. AI tools come fourth, ahead of travel brands and journalists. The machines are rising, but they don't beat people just yet.

And AI disappears at booking. Although AI is widely used for inspiration and research, there is a credibility gap at the most sensitive moment of the transaction.

For now, giving an AI agent a credit card is too much to ask for, but many could be persuaded. Asked what would make them trust AI to book, 52% of travelers want to review and edit before confirming, and 30% want a proven track record. And that history of success might come sooner than we expect, with the viral success of AI agents like Instinct and Muse.

Commerce: Agents are coming for one of travel's biggest pools of dollars

Instinct, the text-based AI agent, may be pacing at nearly $500 million a year in travel-booking volume. That's based on the founder saying the app is running at least $1 billion a year in transaction volume, and 50% of that is travel. But, in the grand scheme of things, that's small potatoes. How big is the opportunity? We estimate hotels pay about $30 billion a year in direct distribution costs and about $75 billion in indirect, so $105 billion-plus. If agents took just 10%, that's a $10 billion-plus opportunity.

But, as a philosopher once said, "mo' money, mo' problems." Agentic bookings are not a free lunch. My colleague, Sean O'Neill, reported that "the P&L doesn't lie", noting that hotels are seeing more AI costs than bottom-line gains. Current user interfaces are also not designed for agentic workflow. 

We're in a sticky phase. Agents are here but most travel-booking platforms aren’t ready for them. It's going to take time, money, and energy to rebuild travel-booking rails for agents. Still, it feels increasingly clear in the past few weeks that the future of travel commerce is here and it will look like agentic AI.

Operations: If you stand for nothing, what will you fall for?

We asked travelers whether they'd switch brands if AI found a better deal. 55% said they were likely or very likely to. Loyalty is a core part of most travel commercial strategies, but that may be at risk as AI nudges travelers to leave the well-worn channels. That's an opportunity for new brands and a threat for incumbents.

Part of the risk is brand sameness: 74% of U.S. travelers agree that many hotel brands feel very similar. French Polynesia invented the overwater bungalow resort. But now Sandals has them in Jamaica and Hilton has them in the Maldives. 

If you can't tell these experiences apart, how is AI supposed to? Brand loyalty has become an operational challenge. How do you create an experience that stands out in an AI world?

Experiences: Fewer results, same destinations

We're moving from a search-led world to a curation-led one. The old way: search, scroll, compare. The new way: ask, shortlist, decide. With AI there is less shelf space. When I asked Instinct for a flight, it gave me one option. One.

In our LLM study for The Mystery of the Disappearing Destination: How AI Is Transforming Travel Discovery, AI sent 38% of recommendations to the same five destinations, and 48% to the top 10. Then again, the top five most-visited destinations get 40% of global arrivals, and the top 10 get 56%. So AI is, if anything, slightly less concentrated than the world it's reflecting back.

Perhaps even more dangerously, we measured how AI tools “pitched” destinations to prospective travelers. Price (21% of themes) and connectivity (16%) were the most common selling points. That's a pretty generic way to recommend a destination.

It's likely that AI is recommending the same few destinations, and, even if you are lucky enough to be included in the consideration set, your destination is being sold as interchangeable rather than on its merits.

The challenge is to build authentic, meaningful, connected, and transformative experiences that give AI surface area to grab on to. Without them, it recommends you in a totally undifferentiated way.

Have you booked using an AI agent yet? How did you find the experience? Let me know your view on how travel commerce will evolve — as always, it’s [email protected].

Best,

Seth Borko

Which AI bets actually move price, demand, and margin, and which ones just demo well?

The CEOs of KAYAK and Skyscanner, IAG's Chief AI Scientist, and Accor's SVP of Data Science & AI, in one room of 200 senior travel executives.

Sania Lali sits down with Benjamin Cany, Vice President of Product and Business Strategy for the AI Delivery Hub at Amadeus, live from the Skift Data & AI Summit in London.

They explore what it actually takes to move AI from a promising pilot into a live operational environment, and why trust, not technology, is the real bottleneck.