Airlines are still at the early stages of agentic artificial intelligence (AI). At the Elevate conference in Atlanta, ATPCO revealed that 37% of carriers have active AI programs but only 11% of airlines have deployed agentic AI.
United and Malaysia Airlines are two carriers that have adopted agentic.
In early September, United Airlines added agentic disruption management to its app, automatically rebooking passengers if their flight is canceled and putting them on standby for up to three earlier flights. If the seat becomes available, passengers can choose to switch to the earlier flight.
Malaysia Airlines launched its Mavis AI assistant in early 2026, based on Ada’s agentic customer experience model. In addition to handling natural language queries, Mavis can handle even complex passenger requests, handing off to human agents only where necessary.
While airlines are dipping their toes in agentic AI, its use for bookings seems further away, even as Phocuswright research shows more travelers are open to letting AI book for them.
One reason for airlines’ slow reaction is a lack of clarity surrounding whether travelers will embrace end-to-end agentic bookings and, if they do, whether airlines will be the ones to carry out these bookings rather than online travel agencies (OTAs) and other intermediaries.
Research firm TNMT, part of the Lufthansa Innovation Hub, is trying to kickstart the process by setting out its vision in a report, How agentic AI could reshape flight booking in the 2030s, which is intended to help airlines prepare.
The report draws on trends in technology, regulation, consumer behavior, geopolitics and airline distribution and envisages four possible futures.
Scenario No. 1: AI search but no agentic booking
AI-powered search in travel is soaring, but the TMNT report outlines a future in which consumers refuse to hand over the final booking decision to an autonomous agent.
If this pans out, airlines will need to remain focused on direct channels such as airline.com and mobile apps, while fares, seat maps, ancillaries and loyalty benefits are only partially accessible in a format AI agents can reliably use. Indirect sales overtake direct bookings, and look-to-book ratios soar as a result.
Richard Edwards of airline strategy consultancy Revidio said that even the definition of an agentic booking is unclear—is it truly an end-to-end process including payment or just up to the point where the user has to agree and choose a payment method?
“No sensible person is going to give up their own financial controls 100%,” he said.
“In a corporation, you also have a chain of approvals. Even if you've given the agentic agent the authorization to go ahead and pay, I would expect to still be given the option to say I agree with that or not.”
John Morhous, chief experience officer for Flight Centre Travel Group’s corporate brands said, “The reality is that in B2B, the company ultimately pays, so I don’t see things evolving to a place where that full decision-making authority is handed over—the human is accountable, not the agent.”
Scenario No. 2: Europe loses access to the best AI models
The report’s second scenario envisages a world where Europe is cut off from the world’s leading foreign AI models, such as when the U.S. government recently ordered Anthropic to restrict Fable and Mythos. In this scenario, European airlines might try to respond by securing long-term access to European computing infrastructure and building a shared, airline-controlled offer layer.
Ann Cederhall, co-founder of airline distribution consultancy LeapShift, is not sure this will happen.
“Kudos if you want to do this kind of European effort, but where's the money? It's going to take 20 years just in Germany,” she said.
Revidio’s Edwards thinks the International Air Transport Association (IATA) would have a logical role in such a plan but it has already shown that standardization, such as in the case of New Distribution Capability (NDC), takes a very long time and risks intentional delay tactics by certain stakeholders.
“There are certain participants whose business is likely to suffer as a consequence of NDC. Some of the guys that participate in standard setting are just in there trying to create longevity for their business model.”
Scenario No. 3: Airlines are shut out
A third scenario from the report sees travelers starting to trust AI agents to make decisions and complete bookings on their behalf. However, in this potential future, airlines fail to secure the customer context and transaction layer because they remain relatively invisible within AI at the expense of indirect channels such as OTAs.
Flight Centre’s Morhous believes airlines will not take full control with agentic and that the winners will be those players who hold “the most complete, policy-aware version of the offer.”
“Our approach is to own both the agent and the plumbing—that is why we built our own AI platform, and AI agents on top (Sam, Mel), and are enabling access to them via MCP. That lets the traveler use an auditable and permissioned way to ask for something, instead of an airline’s myopic options,” he said.
Revidio’s Edwards said OTAs already have first mover advantage in agentic AI.
“[OTAs] have seen that an opportunity to disintermediate again and to compete against those airline direct sites. They have direct connections into the hearts of many travel businesses—airline, hotel, car and rail. They see themselves as being in a very, very good position to quickly plant their feet in this new territory.”
While airlines have been slow to react, he does think they will recognize the danger posed by agentic AI, but this may take three or four years.
Scenario No. 4: Airlines embrace agentic
In the final scenario, airlines have recognized that travelers want to book agentically and have trust in AI agents. They also choose to make their full offer visible, understandable and bookable inside those agentic environments.
Gaurav Roy, CTO of airline software provider Accelya, said, “The strategic question for airlines is whether they want to participate as intelligent retailers or they are just an inventory provider in this whole ecosystem.”
Roy believes airlines will be forced to up their data and systems whether they want to or not.
“Others are going to sweep them anyway. If airlines don't expose their data to AI, AI either hallucinates or barrages them to get some information out of them,” he said.
He added, “Airlines certainly want to own the context and getting information about the purpose of travel saves a lot of infrastructure cost at the airline because you're not putting irrelevant offers out there.”
Which scenario will win out?
TNMT points out in its report that the goal was not to predict one definitive future but rather to “explore what each could mean for airline distribution, customer relationships and commercial strategy.”
LeapShift’s Ann Cederhall predicts that One ID, IATA’s digital identity initiative, will help agentic AI grow in airlines.
“I believe that I should own this [identity and payment] data and this can be controlled by my own AI agent in a beautiful way. I don't see an agentic AI agent being really any different from a travel agent, except that it's going to be much more efficient.”
Accelya’s Roy said trust will be key to travelers adopting agentic and airlines being forced to follow.
“The journey of trust, for some people, takes a few weeks; for some people, it takes a few years.”