The corporate booking tool may not disappear in the age of artificial intelligence, but travelers could see much less of it.
That was the broad view among participants in a BTN technology forum in London examining how AI could change the form, functionality and governance of corporate travel booking. Panelists envisioned booking capabilities embedded in employees' existing work environments, where an AI assistant might identify the need for a trip, recommend an itinerary and initiate the booking process.
Behind that new interface, however, much of the machinery of managed travel remains. Supplier content still must be sourced and serviced. Corporate policy still must be applied. Bookings still require approvals and audit trails. Travelers still need support when a trip becomes complex or something goes wrong.
"The booking tool will not be visible," said Havas travel manager for the UK Roberta Iorizzo. "It will probably be still there," but travelers instead could speak to a mobile device, describe what they need and receive proposals or recommendations, she said.
FCM Travel chief customer officer Jo Lloyd similarly predicted that the underlying technology would remain even as the traveler's means of accessing it changed.
"The tools will be there. The plumbing, the pipes, those will still be in play," she said. "But the way that the community interfaces with booking tools I think will be very different."
She offered the example of an employee placing a meeting in New York on a calendar and receiving a prompt asking whether the traveler wanted to book their usual flight.
Evolution Beneath a New Interface
The industry is cautious not to characterize AI as a wholesale reinvention of travel booking. Some of its most immediate effects may be to improve or accelerate processes the industry already performs.
Sage Consulting principal consultant and technology specialist Guy Snelgar pointed to reshopping as one example. AI could make such processes more efficient and improve the filtering of relevant fares, but the general function is not new.
"It won't necessarily be revolutionary," he said. "It will be doing things that we're doing today, but maybe slightly more efficiently."
That distinction matters. A conversational interface may make the booking experience appear radically different, but it does not eliminate the commercial, operational and technical structures beneath it. Bookings still may run through global distribution systems or direct connections. Policies and approvals still must be enforced, and travel management companies still must service the resulting reservations.
Even agentic booking does not necessarily change every line of accountability, according to Lloyd. An AI agent may be a new way of consuming information and initiating a transaction, she said, but the booking ultimately is still being made through established systems.
The risks become less clear, however, when one AI initiates an instruction that another AI executes. A calendar assistant, for example, might interpret a meeting invitation as requiring travel even though the employee plans to attend remotely.
Booking Moves Inside the Enterprise
The more consequential change may be where the traveler encounters the booking capability.
Rather than directing employees to a separate booking application, corporations increasingly may seek to bring travel shopping and booking into chat, calendar and enterprise AI environments.
"It's going to come through the chat," said Travlr ID founder Gee Mann. "It's going to find itself inside the enterprise firewall."
Model Context Protocol, or MCP, is one potential mechanism for connecting enterprise AI assistants with travel data and services. Mann described MCP as "a USB port to your data inside your business," allowing an authorized AI system to access and interact with information from connected sources.
For a traveler, that could mean asking an enterprise assistant about an upcoming trip. For a manager, it could mean posing a conversational question about travelers who have not completed a required action, rather than opening a business intelligence dashboard, exporting a file and distributing a report.
Corporations also see a broader strategic benefit in keeping employees within their own controlled environment.
"Our enterprise customers, they don't want any of their travelers to go outside of their own ecosystem," Lloyd said.
That approach could give the corporation greater control over the employee experience and potentially more direct access to behavioral data. It also could weaken the booking provider's control over how its content, policy configurations and recommendations are presented to the corporate user.
Content Still Comes First
However sophisticated the interface becomes, the booking experience will succeed only if it contains the fares, rates and options travelers expect.
For Iorizzo, content remained more important than the booking fee or the novelty of the AI.
"The main problem for any program is people going out of the program," she said. Companies need to eliminate the familiar traveler complaint that a better fare is available directly from an airline or through a consumer booking site.
AI does not by itself aggregate airline or hotel content. It may, however, accelerate the coding and development required to establish direct connections, incorporate additional content and alter booking flows.
"The speed at which these things are changing is unlike anything we've ever witnessed before in our industry" said Lloyd.
Yet buyers still need to distinguish between what has been demonstrated and what remains on a provider's roadmap. After evaluating multiple TMC and booking-tool proposals for an active sourcing effort, Iorizzo said she had not seen a compelling demonstration of AI bringing a previously unavailable breadth of content into the booking experience.
From More Content to Better Recommendations
A larger content pool creates its own problem: deciding which options to show.
An AI-enabled booking experience cannot simply retrieve every available flight or hotel. To reduce complexity, it must determine what is relevant to the corporation, the trip and the individual traveler.
That requires more than a traditional traveler profile.
Static profiles generally contain fixed information that can be passed through existing travel systems, such as loyalty details and basic preferences. Dynamic profiles could incorporate behavior, context, changing needs and information relevant only in certain circumstances.
"A dynamic profile will learn, but also confirm," Mann said. "A static profile would just state."
The distinction is significant because a stated preference is not necessarily universal and often is not supported by actual traveler behavior. A traveler may favor one option on an individual trip and another when traveling with colleagues. Preferences also can change over time or become irrelevant in a particular context.
Dynamic profiles therefore must work beside static profiles rather than simply replace them. Existing systems may accept only a limited set of fields, while newer booking tool technologies already claim to use a broader range of contextual or accessibility information to create more relevant recommendations for travelers.
Most booking tools also claim they will “balance” travelers’ stated and revealed preferences alongside corporate policy. Again, this is a dial turn that will need to be stress tested by program administrators to understand how nuanced and how corporate-friendly the configurations can be.
Who Controls the Memory Layer?
And while the ability to recall traveler behavior could become one of the most valuable parts of an AI-enabled booking environment, it will also be one of the most sensitive.
"Just because you learn it doesn't mean I want it," Mann said. An AI may infer a preference or personal attribute from past behavior, he explained, but it should confirm that inference with the traveler rather than treating it as a permanent fact.
That raises questions about not only privacy, but ownership and control. The traveler may want certain information used for one type of trip but not another. And that’s just at the individual level.
A memory layer combined with AI may have the ability to deduce business strategies based on travel patterns. How that information might be published through other corporate or partner systems is hard to trace, and companies have voiced concerns, despite the fact that they also want to lean into a personalized traveler experience for their program.
Snelgar described a client's reaction after learning that a widely used corporate booking and expense system had enabled as a default setting the use of the client company’s travel data for training. The company switched the training function off rather than devote extensive resources to investigating how the data might be used. Legal and security teams may go as far as to restrict information sharing even where it could improve the traveler experience, depending upon the organization.
"If you don't control the data and somebody else is training the AI, they're getting the insights, they're getting the recommendations," Mann said.
Auditability Is Not Ready
While most of the booking tools covered in BTN’s 2026 guide have stopped short of executing transactions without a human touchpoint, the capability to do so is available. As AI advances beyond recommendations, the ability to reconstruct its decisions becomes critical.
A routine booking error may result in a cancellation or additional cost. A failure involving duty of care could place a traveler in an unsafe location or prevent a company from responding properly during a crisis.
Asked whether technology providers currently could produce a satisfactory audit trail showing how an AI reached a decision, Mann was direct: "The answer is no across the board."
Snelgar said providers ultimately will have to document the authorization behind a booking. Some corporations already may refuse to pay for a ticket issued without a record of the required approval, he noted. Agentic booking will make that traceability more important, not less.
As AI advances to human-free transactions, how providers configure systems to pause for interventions will be critical. And they will need to demonstrate that the AI does not override its own configurations.
"There will be a need at some point for the AI to realize when it needs to stop," Iorizzo said, citing complex trips as well as crisis situations.
A New Booking-Tool RFP
While many of the underlying mechanisms will remain for booking tools, AI promises to bring profound changes in how bookings are executed and how much human touch—or even human awareness—will go into making those bookings. To that end, booking tool procurement can no longer concentrate primarily on content, geographical coverage, user experience, service and transaction fees to be comprehensive.
Buyers also need to ask how providers train their AI, whether corporate or traveler data is used for that training, what information is retained and how travelers can correct or delete inferred preferences. They need to understand the boundaries of an AI agent's authority, the human escalation process, and whether the provider can document the data, policy and approval behind each action.
Travel teams are likely to require deeper participation from legal, information security, data governance and corporate AI leaders.
"You need to get your data processing teams to really dig down on your suppliers," Mann said. "Start from scratch and ask fresh questions."
Procurement becomes still more complicated when the booking interface sits inside an environment controlled by the corporate customer.
"When it comes into your house, your provider is no longer in control of that experience. You are," Lloyd said. "Once it comes into your environment, we lose sight of it."
That shift could divide accountability across the provider supplying the travel content, the system executing the transaction and the corporation controlling the AI interface.
Many Questions Yet to Answer
The forum began with a question about the future form of the online booking tool. It ended with much larger questions about enterprise technology governance.
Who owns the dynamic profile and its memory? Who determines which inferences are appropriate? Who is accountable when one AI gives another an incorrect instruction? Can a provider produce a complete audit trail? Does responsibility shift to the corporation when booking is embedded inside a customer-controlled enterprise assistant? And how should the industry price an experience involving continuous searches, AI interactions and automated servicing?
The interface may be the first and most visible part of corporate travel booking to change. The harder work will be determining who controls, governs and accepts responsibility for everything that happens behind it.