What Does the Future Hold for Agentic Commerce in the Travel Industry?

November 2026
Fintech & Payments

Mastercard Preparing Travel Agentic Commerce for Takeoff

On the 12th November, Mastercard announced it had successfully completed its first live end-to-end AI-powered travel booking and payment transaction, in collaboration with Travelport, Booked AI, and Basis Theory. Travelport is a UK-based travel technology company that primarily connects airlines, hotels, car rentals, and rail companies to travel agents. Booked AI is an Australian early-stage startup that runs a conversational AI travel agent. Basis Theory is a US-based payment infrastructure company offering a processor-agnostic payment vault; positioning it as a foundation for agentic payment flows. 

This process saw Booked AI’s AI travel agent source a live New Distribution Capability (NDC) offer from a major airline through Travelport’s TripServices. Once the customer confirmed the flight and fare details in the conversation platform, Mastercard Agent Pay enabled the agent-initiated payment using a secure tokenised credential, with the underlying card credential held in Basis Theory’s PCI-compliant vault. After the payment was authorised, the booking process went through Travelport’s and the airline’s systems. 

An important part of this is the fact that the consumer had to confirm the process at multiple stages. The consumer’s input was needed to approve the booking and the amount paid. Further securing the payment was Mastercard’s Verifiable Intent, which linked the approval to the transaction; allowing the payment to be checked against the customer’s original instructions. This is important as a significant concern around the use of AI agents is the potential for an agent making a purchase the consumer did not wish it to, which complicates dispute and chargeback processes on the airline’s end. By gaining the consumer’s approval through these steps, the consumer cannot reasonably claim a transaction was made against their wishes. 

Another important factor in this transaction was the use of tokenisation. By providing the AI agent with a token, but never giving it access to the actual card details, the process is made more secure. AI agents having access to an individual’s payment card details is a significant risk, as this would make them the prime target for criminals looking to steal the information. There is also a concern over whether an AI agent could be tricked into providing stored payment details in the same way that other AI chatbots have been tricked into giving responses to which they were not intended. Both these concerns are alleviated by using payment tokens instead of giving agent consumer’s payment credentials. 

In the press release about this achievement, Andrew Jordan CTO at Travelport focused on the continuity for the airlines. The airline receives the payment on the same rails as a current purchase made through a traditional travel agent. The airline remains the merchant with no intermediary checkout. This is important for the airlines as it allows them to maintain the current control they have over the process. 

Do Airlines Want to Accept Agentic Commerce?

It is important for those looking to pursue AI agent-powered travel purchases to consider airlines’ concerns, particularly as parts of the industry are making moves to prevent AI agents becoming a key part of the travel market. Most notably, on the 11th November, United Airlines announced that it was restricting access for unauthorised third-party AI agents. In its statement, United Airlines raised customers’ data security, the risk of unauthorised purchases, use of loyalty rewards, and potential compliance issues with bookings made by AI agents. Mastercard’s method addresses the concerns around giving consumer payment credentials to the agent and the risk of unauthorised purchases. It should also be relatively simple to enable the agent to gain the consumer’s permission to use airmiles or not. The regulatory concerns are less easy to address; especially the need for a human to acknowledge their understanding to rules around what they can bring in luggage. It is important to note that there is nothing to suggest that this United Airlines ban is related to Mastercard’s tests. 

It is also possible that United Airline’s decision is not only based on security and compliance concerns, but also on commercial interests. If consumers increasingly use AI, this will reduce the human consumer traffic from the airline’s own websites and apps; limiting the amount of useable data they have access to from these sources. It will also make it harder to use consumer behaviour to adjust ticket costs; removing the ability to make customised offers or maximise revenue from an individual. With airlines’ margins being so small, airlines have made a concerted effort to encourage direct booking over travel agents. This has only been exacerbated by global fuel price rises. Despite this, the use of NDC booking reduces the cost for the airlines; meaning if AI agents use NDC, this could offset the cost of bookings being made through a travel agent. 

There could also be concerns that the established airlines cannot utilise their reputational advantage over smaller carriers when their being compared by an AI. While United Airlines has not stated any of these as motivating the move to ban unauthorised third-party agents, they remain potential challenges a major airline could face as third-party AI agents become more popular for flight booking. 

It is worth noting that United Airlines has left the door open to authorised or in-house AI agents. Using such agents would allow United Airlines to ensure its consumers are using secure agents, while also having more control over which airlines are shown to the consumer. 

What Do these Developments Tell Us About the Wider Market? 

These developments show that players in the travel space are taking agentic commerce very seriously. Due to the complexity of booking travel, needing to book different legs of a journey while also covering accommodation, insurance and other factors, AI seems like it has strong potential to streamline this process for consumers. The challenge for AI travel agents is getting buy-in from the different parties involved. Juniper Research expects that this space will become the primary domain of specialist AI travel agents, with most major service providers not allowing access to generic chatbots to make purchases. Over the medium term, this will be a mix of all-in-one third-party agents that manage the whole process, and specialist agents that focus on one part of the market or built into an existing travel service.

Payment service providers (PSP) must look to enable AI agents to initiate payment on behalf of consumers without the agent having access to the users’ actual payment credentials. Tokenisation is a strong way to approach this. It is vital that PSPs ensure that final approval remains with consumers to prevent liability issues over unauthorised purchases. 

Latest research, whitepapers & press releases

Got a question? Try our AI assistant