Why AI Travel Planners Can't Actually Book Your Trip

A traveler asks ChatGPT to plan a long weekend in Tokyo. Within seconds, they have flights, hotels, and restaurants. Then they try to actually book something. Suddenly the system breaks down, the AI cannot complete the transaction, and the traveler is stuck toggling between tools.

This is not a fringe problem. It is the central crisis facing the global travel industry right now.

Consumers are moving to AI planning tools faster than travel platforms can handle. The gap between what AI promises and what legacy booking systems can deliver is costing the industry real money, real customers, and, increasingly, real business value.

The Infrastructure Cannot Keep Up

The backbone of global travel bookings are Global Distribution Systems, known as GDS platforms. Think of them as the central nervous system connecting airlines, hotels, and travel agents. They were built decades ago, long before anyone imagined travelers would be typing requests in plain English and expecting a confirmed itinerary in response.

That mismatch is now a competitive liability. AI-sourced traffic was 47% less likely to convert into a completed booking in mid-2025 compared to traditional search. That figure had improved from 86% just nine months earlier, which shows the problem is closing but still enormous.

Fixing it is expensive and slow. Large-scale GDS modernization projects typically cost over US$2 million and take nine to 14 months to complete. Most of the industry has simply deferred the work until the pressure became impossible to ignore.

Travelport Makes Its Move

On May 27, 2026, Travelport announced a three-way partnership with Cognizant and Anthropic to tackle this problem head-on. The deal puts Anthropic's Claude AI model at the core of Travelport Trip Services, the platform that handles bookings, exchanges, refunds, and disruption management for airlines, travel agencies, and corporate travel managers.

Anthropic's Opus 4.8 Raises the Bar for Enterprise AI Reliability

Opus 4.8 improves code review reliability and introduces Dynamic Workflows for enterprise teams. Unchanged pricing keeps budgeting predictable.

Mission MediaMission Media

The technical centerpiece is an MCP-based interface layer. MCP, or Model Context Protocol, is a standard developed by Anthropic that lets AI systems talk directly to external databases and booking platforms. In practical terms, it means a traveler's natural-language request can be translated into a confirmed booking linked to live availability, without a human agent bridging the gap.

Cognizant is also deploying Claude inside its own engineering tools to help Travelport's developers work faster through its Neuro AI multi-agent accelerator, which will use Claude for code development, testing, and reviewing changes across Travelport's large, complex codebase.

"AI is not a future consideration, it is happening now, and the companies that move fastest and most intelligently will define the next era of travel technology," said John Mangelaars, CEO of Travelport.

What's Actually at Stake

The business case is blunt. IDC estimates that saving just one agent hour per day across large US travel management companies translates into US$3 billion in annual productivity gains. For travel companies that have already adopted AI, the results are showing up in real numbers: 59% report increased productivity, 30% faster decision-making, and 26% cost reductions.

The competitive threat is coming from multiple directions. Hotel chains including Accor, Hyatt, and Wyndham are building direct integrations into AI platforms like ChatGPT to cut out the middlemen and avoid paying commissions to online travel agencies. That means AI disruption is attacking both the front-end booking experience and the back-end distribution infrastructure at the same time.

Harvard Business Review's January 2026 analysis warned that platforms like Expedia and Booking.com risk losing their gateway role entirely as conversational AI tools answer travelers' questions directly. And in a development that signals how severe the valuation risk has become, Skift reported in May 2026 that AI uncertainty was the specific reason acquirers walked away from buying the world's largest corporate travel company.

The Travelport deal is not a speculative technology experiment. It is emergency infrastructure work, undertaken because the alternative is being made obsolete.