“The question today is no longer whether AI can produce a prototype, but how we can take it further and enable it to truly transform the travel industry.”
At the 2026 TravelDaily Conference, Fabien Blondet, Senior Vice President of Research and Development at Amadeus, shared Amadeus’ perspective on the development of AI in travel in a keynote titled “Creating Value in Travel with AI: From Trusted Data and Ecosystem Connectivity to Deployment at Scale.”
In his view, the travel industry is entering a critical stage in AI adoption. A growing number of consumers are already using AI for travel inspiration and now hope to use it for search, comparison, booking, and even the entire travel journey. According to Amadeus data, 74% of consumers have already used AI tools for travel inspiration, while more than 40% are willing to go further and use AI to book travel.
However, the travel industry itself is highly fragmented. Data silos have long existed across airlines, hotels, airports, car rental companies, travel agencies, and other parts of the ecosystem, making it difficult for consumers to book combined products such as flights, hotels, and car rentals in a single transaction. Although AI can be an important technology for addressing these issues, achieving large-scale adoption requires more than simply adding an “AI chat box” to the front end.
Fabien Blondet believes that for AI to create real value in travel, at least three foundations are required: trusted data, ecosystem connectivity, and the ability to operate at scale.
First, AI Starts Not with Large Models, but with Trusted Data
From Amadeus’ perspective, data is the foundation that enables AI to operate effectively in the travel industry.
Over the past 40 years, Amadeus has continuously accumulated data across aviation, hospitality, and other travel sectors, including flight schedules, inventory, availability, PNRs, orders, offers, and hotel information. Today, the volume of travel-related data within Amadeus systems exceeds more than 30petabytes.
More importantly, this data does not exist in isolation.
Amadeus is using a Data Mesh to break down data barriers among airlines, airports, hotels, car rental companies, and other business areas, enabling different use cases to access shared data and capabilities.
For example, a single flight data model can use consistent data definitions, whether it is used by an airline, airport, travel agency, or Amadeus’ internal IT and operations teams. This unified data model reduces duplicated development and data redundancy, while making interoperability across different business areas easier to achieve.
Fabien Blondet believes that if AI is built on inaccurate, incomplete, or fragmented data, it will struggle to become part of the travel industry’s core business processes. Amadeus is therefore not only building data infrastructure, but also strengthening AI governance, including data security, privacy protection, data sovereignty, and data-access permissions for different use cases.
Amadeus adapts and optimizes AI technologies for travel specific use cases, combining advanced models with trusted travel data, domain expertise and responsible governance.
Second, the Value of AI Comes from Connectivity, Not Just Conversation
Another fundamental challenge for the travel industry is connectivity among its different participants.
Amadeus currently connects a large number of airlines, hotels, travel agencies, airports, and car rental companies worldwide, and its role is inherently that of a connector within the travel ecosystem. Fabien Blondet therefore believes that AI will unlock the greatest value not by simply replacing human customer service, but by becoming deeply embedded in core systems for search, booking, distribution, and operations.
In fact, Amadeus has used machine learning in the search process for more than a decade and is further embedding AI capabilities into search and shopping workflows.
This year, Amadeus also acquired SkyLink, a US-based AI-native company that uses conversational AI to move quickly from travel-product search to booking. The acquisition is also seen as another step in embedding AI more deeply into the transaction process.
At the same time, Amadeus is building its own Travel AI Stack, adding AI capabilities on top of its existing shared systems, connectivity layer, and global infrastructure, while providing APIs designed for AI applications.
As AI agents enter the travel industry, traditional API connectivity is also evolving. Amadeus is exploring emerging protocols and standards, including MCP and other agent-to-agent approaches, that could help AI agents interact with trusted travel content and capabilities. =
In hospitality, Amadeus is also working with Google to explore the Universal Commerce Protocol (UCP), connecting hotel products with a new generation of AI-powered shopping and transaction channels.
In other words, consumers may eventually stop searching item by item through traditional OTA interfaces. Instead, they could communicate their requirements directly to an AI agent, which would then call different airline, hotel, car rental, and other systems to assemble the complete travel offering.
This also means that travel companies need to move gradually from “designing interfaces for people” to “providing products and services that AI agents can access and invoke.”
Third, the Real Challenge for AI Is Moving from Demonstration to Scale
For the travel industry, getting an AI prototype to work is not especially difficult. The real challenge is placing it in a production system that handles millions of transactions, or even more, every day.
Fabien Blondet said that Amadeus systems currently support enormous volumes of travel transactions every day, peaking at one million transactions per second, including more than 150,000 requests per second coming directly from customers and travelers.
As AI agents gradually become a new gateway to travel, Amadeus expects traffic from AI agents over the next five to ten years to reach 10 to 100 times the level of traditional traffic.
This means that AI must not only be “smart”; it must also be sufficiently stable, fast, and cost-efficient.
General-purpose large language models still face challenges such as high costs, long response times, and hallucinations. For travel infrastructure that must operate 24/7 and meet 99.99% availability requirements, these issues, when magnified across massive transaction volumes, could directly affect the entire system.
Amadeus is therefore optimizing general-purpose models for travel use cases. Through model fine-tuning, LLM caching, front-end optimization, and other approaches, Amadeus aims to reduce AI invocation costs and response times while improving the consistency and reliability of outputs.
Amadeus has also migrated all of its traffic to Microsoft Azure and entered into a strategic partnership with Google Cloud, with plans to deploy additional Google Cloud processing capacity in Asia and the United States. Through a multi-cloud architecture, Amadeus aims to improve system stability and flexibility while gaining access to the latest AI infrastructure capabilities.
In addition, Amadeus is collaborating with technology companies including Databricks and Anthropic to integrate cloud computing, big data, and large-model capabilities more deeply into the travel technology ecosystem.
Ultimately, AI Will Transform the Entire Travel Value Chain
In Fabien Blondet’s view, the ultimate value of AI is not limited to “chatting with travelers” or “answering questions on behalf of customer service teams.”
Amadeus aims to use its capabilities across the entire travel value chain to bring AI into multiple stages before, during, and after a trip.
For example, in corporate travel, AI can use traveler behavior and demand signals to generate personalized product recommendations in advance. When a flight is delayed or canceled, AI can connect flight, hotel, and other systems to automatically identify alternatives for the traveler, rather than simply stating which flight departs next.
In aviation, Amadeus is also using AI to help airlines create and manage product catalogs, enabling them to design and sell different product combinations more flexibly.
In hospitality, AI can help hotels present destination information and details of nearby attractions more efficiently.
The common thread across these use cases is that AI is no longer merely a standalone tool. It is beginning to enter the existing core systems of travel companies.
This is also central to Amadeus’ view of AI development in travel: the real value of AI lies not only in generating content, but in connecting data, products, systems, and transactions.
Only when trusted data foundations, open ecosystem connectivity, and sufficient technological scale are in place can AI evolve from a “new technology tool” into infrastructure that improves efficiency and drives changes in travel business models.
For travel companies, the question that truly deserves attention in the next phase may no longer be whether to use AI, but whether their data, systems, and products are ready to be connected and accessed by AI.
