Give an AI assistant a vague request to arrange a trip to New York, and it still has to produce an itinerary precise enough to book.
“It can’t drop you in Philly and say, well, we’re close,” Andrew Jordan, chief technology officer at Travelport, told Newsweek. “It has to be New York, on a day you wanted, at the time you wanted.”
Travelport, which connects airlines and hotels to travel agencies and booking platforms, is working with Cognizant and Anthropic to use Claude in developing and testing its software. The collaboration initially focuses on TripServices, Travelport’s booking and servicing platform, which launched in June.
Jordan discussed the work as part of the Seeking Value video series presented by Cognizant. Joining him was Nambi Palanivelu, vice president and strategic business unit head at Cognizant, overseeing retail, consumer goods, travel and hospitality. Gabriel Snyder, Newsweek’s executive editor, enterprise, moderated the conversation.
...
AI assistants can help travelers explore destinations and compare options, but completing a reservation requires specific dates, availability and booking conditions. A request to book a trip within company policy, for example, leaves open whose policy applies and which restrictions govern the booking. Those ambiguities have to be resolved before a reservation can be confirmed.
Palanivelu said the demand for precision applies to hotel reservations, payments and loyalty-point redemptions.
“The AI can reason and recommend, however, the system of record has to commit,” he said.
Jordan believes consumers have another reason to hesitate before handing purchases over to AI: trust. He compared the prospect to autonomous vehicles, arguing that people’s willingness to surrender control can lag behind the technology itself.
“The minute you’re actually committing your money, that trust bar just suddenly raises quite significantly,” he said.
The stakes rise when travel plans unravel. Jordan identified unexpected disruptions, including delayed and rerouted flights, as the hardest part of servicing trips with AI.
Mistakes in handling refunds can expose to significant regulatory penalties.
Cognizant is applying Claude to the engineering work behind those transactions. Palanivelu said the models can trace business processes through existing software to generate test cases and update those tests as policies and data structures change.
Consider a canceled flight affecting 200 passengers. Some travelers may need to be rebooked, others may require partial refunds and the transactions could involve multiple currencies, jurisdictions and time zones.
“All of these scenarios cannot be written by engineers or testers,” Palanivelu said.
Claude can generate additional tests to account for such complications and check AI-generated code against existing business rules. Palanivelu emphasized that engineers and testers still make the final decision about what enters production.
The work also requires understanding the older software major travel systems depend on. Decades of development can leave companies with complex codebases whose original designers are no longer available to explain how the software works.
“Legacy code means business rules that were never written down,” Palanivelu said. “Now AI, for the first time, helps us write it down.”
He described using Claude to follow business functions through millions of lines of code, translate embedded logic into language that travel specialists can understand and preserve existing system behavior as a set of tests. Engineers can then decide which portions of the software to retain, modernize or leave untouched.
Jordan described how the collaboration could help Travelport improve its existing technology and develop new products.
"It's a new engine that has capability baked into it, proven AI models that we can deploy at scale," he said. "And now it's a choice of what we use that engine for."
Travelport wanted an engineering partnership built around AI rather than a conventional outsourcing arrangement that simply added developers to its existing teams.
Cognizant was already a Travelport supplier, and its work with Anthropic gave it experience with the AI models Jordan wanted to bring into the company’s engineering process. Palanivelu said Cognizant combined its engineering resources and travel-industry expertise with Anthropic’s models to demonstrate elements of Travelport’s plans before the companies committed to the larger effort.
The pace of AI development reinforced Jordan’s decision. Software releases once followed schedules measured in months or years, giving companies time to prepare. New frontier models now arrive much more frequently, presenting businesses with a steady stream of capabilities to evaluate and potentially incorporate into their products.
“I don’t think that anyone, big or small in the world today, can credibly say that they can solve that problem on their own,” Jordan said.
Working with Cognizant gives Travelport access to experience gained from deploying AI models more broadly, Jordan argued, making it easier to recognize which advances could improve its internal operations or customer-facing products.
Jordan said AI assistants could change the practices that have guided online retailing for decades. Travel companies have traditionally relied on attracting people to their websites through search engines, then guiding them toward a purchase. Someone shopping through an AI assistant may never visit those sites, leaving sellers with less control over how their offers reach customers.
"If you think that you can be successful just playing around with SEO and websites, you will fail very quickly," Jordan said.
Palanivelu noted that retailers must consider how to authenticate those agents, provide accurate product information and handle high volumes of automated requests. Pricing strategies may also need to account for software that can search and compare offers continuously.
Travelport has a particular interest in how those changes develop. Global distribution systems have historically made it easier for travel agencies to access airline inventory through a single connection rather than maintain separate links to hundreds of carriers.
Snyder asked whether the growing use of AI agents would increase or diminish the value of intermediaries such as Travelport.
“If it’s done well, it increases,” Jordan said.
Jordan sees an opportunity for Travelport to expand beyond supplying flight and hotel inventory by providing travel agencies with data, analytics and software that help them sell to customers shopping through AI assistants.
The industry is still working out how extensively agents will participate in purchases. Jordan said some airlines have begun limiting traffic originating from AI-driven interfaces, a response he considers temporary.
Looking three years ahead, he predicted that travelers could stop shopping for trips through conventional web browsers. Palanivelu offered an equally ambitious forecast: AI could anticipate flight disruptions 48 hours in advance and arrange a response before passengers need to intervene.
“If AI agents are your buyer, accuracy is your new storefront,” Palanivelu said.