Travelport, Anthropic, and Cognizant: Wiring Conversational Intent to Confirmed Bookings with MCP

The AI rebuild starts on the infrastructure side of travel distribution

Travelport taps Cognizant, Anthropic for AI upgrade

Travelport partners with Cognizant and Anthropic to deploy Claude across travel retailing and distribution platforms.

www.businesstravelnewseurope.com

On May 27, 2026, Travelport — one of the three global GDS operators alongside Amadeus and Sabre — announced a three-way collaboration with Anthropic and Cognizant. A GDS (Global Distribution System) is the backbone infrastructure that distributes airline and hotel inventory to travel sellers worldwide. Under the agreement, Anthropic's Claude will be embedded into Cognizant's engineering operation to modernize how Travelport builds, tests, and maintains the retailing and distribution platforms used across more than 165 countries.

The problem statement in the release is unusually direct. Travelers have already started searching with AI tools, but booking systems built for a previous era cannot interpret or fulfill that intent, and a gap has opened between the two. The three companies frame their answer as "intelligent infrastructure designed to close the gap between AI-driven travel intent and a confirmed booking."

Most agentic commerce coverage — the emerging model in which AI agents compare and purchase on a customer's behalf — fixates on consumer-facing chat surfaces like ChatGPT. What makes this announcement notable is that the actor is not a consumer app but the distribution infrastructure that holds inventory and processes transactions. In retail terms, this is the equivalent of the commerce platform and the inventory and order APIs declaring themselves agent-ready, rather than the storefront.

The foundation: Cognizant and Anthropic's strategic partnership

The lineup did not appear out of nowhere. Anthropic announced a strategic partnership with Cognizant on November 4, 2025, under which Cognizant is rolling out Claude to up to 350,000 employees and integrating Claude Code, MCP, and the Agent SDK into its engineering infrastructure. Financial services was named as the initial vertical focus, which makes travel the next large industry-specific engagement to be made public.

Cognizant CEO Ravi Kumar S has positioned the firm as an "AI builder company" — one that uses AI to rebuild business processes themselves rather than layering tools on top. He described this collaboration as "giving Travelport the tools to move faster and deliver higher quality at scale to meet the challenge of a changing travel distribution landscape." The framing is a long-running transformation of day-to-day software delivery, not a one-off proof of concept.

Claude's first job is assisting code development, test creation, and pull-request review. Using its long context window, it reads across Travelport's codebases and surfaces the business logic embedded in years of accumulated code. Rich O'Connell, who leads alliances at Anthropic, put it plainly: "Reasoning across large, complex codebases is where Claude is at its best — and that's exactly what travel infrastructure demands."

First target: Trip Services, the booking, exchange, and refund platform

The initial focus of the collaboration is Travelport Trip Services, the API platform that handles bookings, exchanges, refunds, and post-purchase servicing, built on the company's cloud-native platform. According to Travel Distribution News's March 2026 analysis, API-driven transactions have grown from 43% of Travelport's volume in 2022 to 63% today, making Trip Services the spine of the business.

On the customer-facing side, automated exchanges and rebooking plus disruption intelligence are named as the first capabilities, surfacing options that account for cancellation and delay risk in real time for travel sellers. The release argues that for a TMC — a travel management company that runs corporate travel programs — saving even one hour per agent per day translates into millions of dollars in annual productivity improvement. The first customer-facing capabilities are expected to reach market within 2026.

Viewed through an agentic commerce lens, it is telling that the first commercialization targets post-purchase work rather than discovery or promotion. Exchanges, refunds, and disruption handling are repetitive yet cognitively heavy, which makes the value of automation easy to measure. These are the travel equivalents of cancellations, returns, and delivery-problem handling in e-commerce, and the sequencing — agents landing after the sale before they land before it — is directly transferable.

An MCP layer that turns conversational intent into confirmed bookings

The most strategically loaded element of the announcement is Travelport's commitment to MCP (Model Context Protocol) as its upper interface layer. MCP is the open standard Anthropic released in November 2024 for connecting AI agents to external systems and data, and it was transferred to the Linux Foundation in December 2025. OpenAI, Microsoft, and Google have all adopted it, and according to Skift's explainer, travel players including Amadeus, Booking.com, and Expedia are moving on it as well. For a broader map of the connection standards, see our overview of MCP, A2A, and UCP.

The release states that this MCP-based architecture "will allow conversational traveler requests to translate directly into confirmed bookings with live availability." CEO John Mangelaars explained the choice of Anthropic in terms of both the protocol's origin and trust.

"

Anthropic also developed MCP, the protocol that lets AI agents interact directly with external systems and data. Choosing the organization that invented that protocol was a straightforward decision. Their approach to safety, reliability and controllability matters as much: travel is a high-trust environment where data is sensitive and the consequences of errors are real.

Source: John Mangelaars, CEO of Travelport

A connection standard alone, however, does not make commerce. Rival Amadeus argued in a May 2026 blog post that MCP is only a first step and cannot by itself carry the complex shopping, booking, and servicing workflows travel requires. Travelport's design reads as one answer to that limitation: the Trip Services transactional fabric sits underneath, the MCP interface layer sits on top, and the two together are meant to run a single line from conversation to confirmed transaction.

The scope still needs to be read precisely. As of the announcement, the conversational booking layer has not shipped, and nothing has been published about how payment delegation or identity verification would be handled. What is described is a connection between intent, inventory, and bookings — not an autonomous purchasing agent that completes payment on its own, and that distinction is worth keeping.

Three layers of labor, and a redefined GDS

Sorting the three companies into layers shows how an AI-era commerce stack is assembled.

Behind this structure sits Travelport's own repositioning. The company now describes itself not as a GDS but as "the infrastructure layer that makes AI-driven travel commerce possible." Travel Distribution News, citing the $50 million shareholder investment closed in March 2026 and 12% EBITDA growth in 2025, concluded that "the GDS era may not be ending. It may be transforming into something else entirely" — and that Travelport is the first of the three majors to make that bet explicit.

Data normalization is the hinge. Trip Services structures content that arrives in incompatible formats — EDIFACT, NDC, low-cost carrier feeds, hotel supply — and the same analysis notes that AI agents cannot work with fragmented, inconsistent data. With Booking.com and Expedia moving on MCP and Amadeus publishing its own standards argument, an infrastructure player that fails to adapt risks being routed around by agent-driven transaction paths. This partnership is best read as a preemptive move against that risk.

Takeaways for retail and e-commerce operators

First, design the conversational entry point and the transaction stack as separate layers. Bolting a chat UI onto a storefront does not close a transaction; inventory APIs, order processing, and post-purchase servicing all have to be callable by machines. In e-commerce, Shopify's Storefront MCP is running ahead on the same thinking, and deciding where to close the gap between discovery and checkout becomes the central design question.

Second, agent readiness starts in the engineering and data layer, not the customer-facing UI. The first place Claude is pointed in this deal is code, tests, and reviews — the same logic as making product and inventory information machine-readable through data readiness work. Modernization that customers never see is what ultimately sets the pace of the features they do.

Third, treat post-purchase operations as the entry point for automation. Travelport's decision to ship exchanges, refunds, and disruption handling first suggests that returns, cancellations, and customer-service workflows are where agentic automation becomes real work in e-commerce too. Building up from areas where value is reliably measurable, before chasing showcase purchase experiences, is an approach that transfers at any scale.

Conclusion

The Travelport-Anthropic-Cognizant partnership is an infrastructure-first agentic commerce case: a travel distribution incumbent pursuing engineering-level AI modernization and an MCP-based interface layer as a single program. The goal of converting conversational intent into confirmed bookings against live inventory is being pursued by rebuilding the transaction stack, not by adding another chat window.

Three things are worth watching from here. Whether the automated exchanges and disruption intelligence due within 2026 pick up real transaction volume. Whether MCP-based booking layers spread to other distribution platforms and OTAs. And which standards eventually cover the pieces this announcement leaves unaddressed — payment delegation and identity verification. The results of this experiment, run first in travel, will arrive in turn for every business that sells complex inventory at dynamic prices.