Where AI Actually Fits in the Travel Experience

There’s a widespread view that AI will revolutionize travel, and no consensus on what that means.

Historically, innovation in travel went to the leisure side first. OTAs effectively replaced the leisure travel agent by providing a better planning and booking experience at a lower cost. Business travel has historically moved more slowly, in part because enterprise requirements, policy, servicing, and integrations make change harder. Only recently have business booking tools begun to approach consumer-grade experiences. This makes sense: consumers can rapidly adopt a new product that outperforms older products. As a result, leisure travel has historically been able to adopt new products more quickly.

Will we see the same here, or will business travel actually lead the way? This depends on where AI is integrated into the travel experience, and whether the interface fits the problem.

Consider a simplified view of the typical traveler experience in September 2026:

The low-hanging fruit is AI trip planning, contextual on-trip messages, and post-trip feedback to help plan future trips. This is relatively easy to implement, and it can be genuinely helpful if the experience is customized to specific travelers. Most importantly, perfection is not necessary for these features to provide value.

It’s worth noting, however, that leisure travel planning is increasingly happening inside the traveler’s ChatGPT or Claude apps, not on the Expedia app or any other OTA website.

For leisure travel, an OTA's personalization features will struggle to outperform ChatGPT or any other personal AI, as those systems can have significantly more context about an individual traveler and their preferences. Those systems don't need to build context from scratch.

This breaks from historical trends, but business travel companies have an advantage here. Unlike Expedia or ChatGPT, business travel companies know not only your own travel history, but the travel patterns of your coworkers. They have more data about the type of trip you’re about to book before you even open their app. Business travelers are already strongly encouraged to use their company’s preferred booking product, and the experience of using that product should theoretically be superior to booking on any consumer product. The data is there, it just needs to be used.

Booking

This one looks the most attractive: Use LLMs to pull structured booking requests out of a conversation with the traveler. They can book a trip just by chatting with your AI, like their own personal travel agent! Travel like a CEO!

The interfaces below are illustrative prototypes. Flight schedules, availability, hotels, and traveler details are fictional.

This is technically possible, and travel companies should, at a minimum, offer this level of service from their conversational AI booking interfaces. However, this only covers a narrow slice of business travel, and an even narrower slice of the overall travel market. For most trips, there is more complexity, more unknowns, and a chat interface may or may not be the best way to book your trip.

There are a few lessons from interactions like this:

• Typing is slow.
• Chat interfaces have a low information density.
• Travelers still hesitate to trust AI, especially with sensitive information.

A chat interface is particularly weak at presenting alternatives, showing the current state of a trip, and letting users revise several choices at once.

In a complex trip with a lot of unknowns, the best interface is one with the shortest feedback loop between your ideal travel plan and the travel options that are available to you.

Everyone wants to travel like a CEO, but CEOs have options. Do not force users into a conversational interface if they haven’t asked for it. If users want to start in a chat interface and quickly switch over to a GUI, that should be supported as well. Conversation is powerful for the right problems, and excellent conversational products know when to introduce richer interfaces or escalate.

Service

This one is the hardest to implement, but it will have the largest impact on the travel industry in the long term. The economics of automated AI customer service make this a question of when, not if, we’ll see widespread automated service operations. In my experience, two things enable a travel management company to win new clients. In order, these are lower costs and a better traveler experience. Clients want to minimize their travel program costs while maximizing the traveler experience, and TMCs that are the most capable of doing this win more bids. AI customer service enables both of these things: charge less than your competitors and treat every traveler like a VIP.

Consider this article from former Citadel Securities COO Daniel Kornum:
AI’s Biggest Winners Have the Lowest Margins.

The argument is straightforward: low-margin businesses stand the most to gain from AI because small reductions in cost enable massive increases in profitability. TMCs know exactly how much money they save if they reduce the average handling time of a traveler-to-agent phone call, and they’re well aware of how much they can save by automating that customer service.

TMCs that are fast to adopt an actually-useful AI here will pull away from the rest of the industry, creating an enormous economic incentive to avoid being the only TMC without an AI service experience. That pressure will likely accelerate partnerships and acquisitions as TMCs with fewer product and engineering resources look for ways to close the gap.

From a technical perspective, service is one of the few remaining parts of the travel experience that still uses a conversational interface. There is enough variation across service platforms, including airlines, hotels, car rental agencies, tour providers, etc., that a conversation has remained the best way to communicate problems and attempt to find solutions. The rest of the travel experience has been saturated with software, and LLMs are causing the same thing to happen to the travel service experience.

What are the commercial implications of travel servicing moving to AI?

The primary first-order effect is obvious: TMCs will become more efficient and able to grow without adding operations headcount at the same rate. The work of operations teams will change as more routine service is automated. Human agents are absolutely still needed, both for complex cases and when AI fails. The goal for service teams will be to reduce avoidable handoffs while protecting resolution quality and customer satisfaction.

For all the hopeful messages about AI increasing productivity without taking jobs away, routine travel servicing is one area where I expect demand for manual work to decline over the next decade. This isn’t new: software has automated repetitive work for decades, and much of travel booking moved to self-service long ago. AI service is a continuation of that trend. Operations expertise will remain critical to designing, supervising, and improving these systems.

The second-order effects get more interesting:

Will AI service become a driver of new profit for both B2B and B2C companies?

As the cost of travel servicing declines, the potential market of customers who want this experience will likely grow significantly. Today, having an on-demand travel agent ready to help you resolve a trip disruption is almost entirely a corporate-travel product. With better experiences available at a lower cost, we'll likely see this expand to the leisure market as well. The need for human agents on standby may decline on the corporate side, while it grows in the leisure sector.

With Altostratus AI, I was up-front with customers that this wasn’t a free service. They could track their flights for free, but the rebooking service would cost money (with an automatic refund if I was unable to help them get home sooner). No one I spoke with objected to paying. The idea of calling American Airlines customer support when you’re stuck in Dallas at 11:30 PM is so painful that travelers are willing to pay to take that pain away. Even after I shut down my company, friends were reaching out asking if I was still running it since they were stuck in an airport somewhere, desperate for a solution.

We may see B2C companies offering proactive disruption assistance as an ancillary service added at checkout, or as an incentive for booking with a specific provider. “Book on Expedia and have your trip monitored by our AI. If something goes wrong we’ll resolve it for you.”

Beyond that, this raises a new question: Who owns service?

If TMCs outsource the technology that powers their customer service, just like some of them are already outsourcing their booking technology, they'll need to differentiate themselves in some new way. This is less of an issue for TMCs building proprietary AI solutions; however, it could lead to disruption among smaller providers with fewer product and engineering resources.

The travel companies that adapt to AI successfully will not force every interaction into a chat box. They’ll use AI where it’s appropriate, chat where it’s genuinely the best interface, and give travelers the best possible experience with the least friction. Once these technologies are widespread, traveler expectations around personalization and service will rise quickly. It's vital for the existing corporate and leisure travel companies to anticipate and build for this future rather than trying to catch up later on.