I believe we are about to see one of the biggest strategic battles the travel industry has seen since the arrival of the internet, because the three major Global Distribution Systems (GDS) are beginning to understand they have exactly what large language models desperately need: accurate, bookable, real-time travel content.
ChatGPT, Gemini and Claude may be extremely clever, but they cannot currently guarantee that a flight is available, apply a negotiated corporate fare, issue a ticket, change a booking or deal with a cancellation. They can understand what a traveller wants, but they still need somebody else to deliver it, which is where Amadeus, Sabre and Travelport suddenly become incredibly important.
The GDSs have spent decades aggregating flights, hotels, rail, car hire and other travel content, while also building the complicated infrastructure required to book and service it. Their weakness has always been the customer experience, as anybody who has worked with traditional GDS screens or corporate booking tools will understand. The LLMs have the opposite problem: they provide a fantastic conversational interface but lack the reliable transactional infrastructure sitting behind it.
Bringing the two together is the obvious next move, and interestingly, the opening moves have already been made
Sabre has formed the clearest alliance by entering a ten-year partnership with Google Cloud, giving it access to Google’s AI technology and Gemini models. Sabre reportedly has around 70 generative-AI projects underway, looking at everything from travel planning and personalised airline offers to customer service and ancillary sales.
Amadeus initially appeared to have aligned itself firmly with Microsoft and OpenAI. It has been working with Microsoft and Accenture to integrate Cytric, its corporate travel and expense platform, with Microsoft Teams, Microsoft 365, Copilot and Azure OpenAI. However, Amadeus has since signed a major agreement with Google Cloud covering Vertex AI and agentic technology, indicating that it does not intend to become dependent on a single LLM.
Travelport has made the most interesting recent move by partnering with Cognizant and Anthropic, the company behind Claude. The stated objective is to connect AI systems capable of reasoning and planning with Travelport’s transactional capabilities. Travelport also operates within the Amazon Web Services ecosystem, which may allow it to access several AI models.
The chessboard is taking shape, although the GDSs are sensibly avoiding becoming entirely dependent on one AI provider.
But it’s the business travel space that is where this game gets really interesting
Leisure travel receives most of the publicity because people enjoy asking ChatGPT to plan holidays, but business travel is much easier to automate because the rules are already in place.
The AI can know the traveller’s preferences, loyalty memberships, company policy, preferred suppliers, negotiated rates, approval limits and payment details. It can also access their diary and emails, which means an employee could simply say:
“Book me to meet John in Frankfurt next Tuesday, get me there before 10 am and home in time for dinner.”
The AI could identify the meeting, search for available content, apply company policy, compare the total journey cost, obtain approval, complete the booking, and add it to the traveller’s calendar. The employee may never need to visit a corporate booking tool or even know which GDS completed the transaction.
This is potentially a much bigger change than simply adding a chatbot to an existing booking system, because the traditional booking interface could effectively disappear.
The GDS may become invisible but more powerful
Airlines have spent years trying to reduce their dependence upon the GDSs by developing direct distribution and NDC connections. However, AI agents create a massive aggregation problem because they cannot realistically maintain separate commercial and technical relationships with hundreds of airlines, hotel companies, rail operators and car-hire businesses.
The GDSs are therefore the logical organisations to aggregate this content, remove duplicated offers, apply the relevant rules and return something the AI can actually book.
Ironically, AI may make the GDS invisible to the traveller while simultaneously making it more powerful within the travel distribution chain.
Where does this leave the TMC?
Travel Management Companies are not going to disappear, but much of their traditional workload will. Simple bookings, amendments, cancellations, refunds, invoices, policy questions, and even disruption management will increasingly be handled by AI agents.
The successful TMC will need to move away from being a booking operation and become the company’s travel orchestrator, controlling policy, negotiating supplier agreements, supervising the AI, managing duty of care and stepping in when a journey becomes too complicated or important to leave entirely to a machine.
The greatest danger is that Microsoft, Google or OpenAI owns the relationship with the traveller, the GDS controls the content and transaction, and the TMC is reduced to providing an invisible support service when something goes wrong.
The crucial battle will therefore be over who controls the corporate travel brain, because that system will decide which content is searched, how company policy is interpreted and why one flight or hotel is recommended ahead of another.
The LLM may appear to be the cleverest piece on the board, but cleverness alone does not win a game of chess. Control of the board does, and the GDSs may already occupy more of it than most people realise.