Standards, not sentences: How to make agentic AI work across travel

Agentic AI has the potential to transform travel by letting conversational assistants move beyond trip research to complete real shopping, booking, and servicing.

For that to work responsibly and at scale, the industry needs shared standards and protocols (especially for interoperability, commerce, and payment) plus high-quality, well-governed data and robust controls.

Artificial intelligence is making information available in previously unthinkable ways, creating fresh opportunities to do things differently across the travel industry.

It's no secret that natural language interactions have been one of the most exciting developments in this process, allowing users new ways to interact with complex data. But it's also important to understand the foundations upon which these conversations are built.

AI, after all, needs protocols to function efficiently. It's standards, not sentences, that will allow AI to grow safely and reliably across the travel ecosystem. Moreover, if we want different systems and AI agents to seamlessly connect, understand, and act on shared data, common standards will be vital.

Here, I want to explore the protocols the travel industry is putting in place to capture the potential of agentic AI, while also looking at the importance of a neutral execution layer, integrated business logic, and a trusted system of record.

Travelers increasingly trust AI-powered systems to research and recommend trips. They may soon allow agentic systems to complete transactions on their behalf. Having grown used to end-to-end, intuitive digital experiences in other areas of their lives, consumers now expect the same in travel.

Many no longer think of a trip as the sum of its discrete parts (flight, hotel, mobility, and experiences) but expect AI to understand and manage the entire end-to-end journey. They expect agents to understand and act upon their personal and historical preferences and intent. They expect communication to be multi-modal (across voice, visual, and text), all while offering always-on conversations and real-time answers.

In this context, it's important to think about how travel sellers can structure, adapt, and expose content to remain relevant in the agentic world. How can they become active participants in those transactions, not just a data feed?

The first step has been to adopt a Model Context Protocol (MCP) approach, an early milestone in AI interoperability. Originally launched by Anthropic and now supported by Microsoft, Google, and OpenAI, MCP provides a universal way for AI agents to connect to tools and access data sources.

However, adopting MCP alone is not sufficient; it is only a first step. It cannot handle complex retailing workflows, such as shopping, booking, and servicing, and cannot provide the end-to-end experience travelers expect. Underlying APIs must also be AI-ready, structured, context-rich, and capable of supporting agent-driven interactions. Important challenges, including latency, usage, and operational costs, also remain. Last but not least, not all use cases require such protocols. For more basic workflows or actions, well-designed REST APIs can often be more than enough.

As developers look for ways to overcome these limitations, commerce and payment protocols are gaining attention. This includes the Universal Commerce Protocol (UCP), which was developed by Google and Shopify (and which is now backed by major players like Amazon, Microsoft, Meta, Stripe and Salesforce) and is designed to support AI-native end-to-end shopping flows.

UCP can complete end-to-end transactions within conversational interfaces and interoperate with complementary protocols like MCP, Agent2Agent (A2A), and Agent Payments Protocol (AP2). This enables AI agents to access inventory, pricing, loyalty benefits, and payment flows.

While UCP could open large new distribution and conversion opportunities as it evolves towards travel requirements, it will also introduce pressure on travel providers to adapt their infrastructure and data quality.

More importantly, they could bear strategic risks, including the loss of control over the user journey and a lack of visibility on platforms’ algorithms.

UCP, then, promises secure, multi-party transactions completed by AI assistants. However, the travel industry poses unique challenges to deployment.

For example, protocols may assume a ‘product catalog,’ which does not fully align with the realities of the travel space. While recent developments are introducing more flexible, live search capabilities, our industry still has no true equivalent to the fixed reference number or SKU (Stock Keeping Unit)-like retail artifact.

For retail in general, this level of orchestration is relatively straightforward. However, for travel, it is far less so. Travel is driven by live search; it is combinatorial, regulated, and service heavy. For example, when a traveler contacts an airline, they receive an offer and an associated Offer ID, dynamically generated based on their request, including origin, destination, dates, fare family, ancillaries, and disruption rules. Servicing flows are even more complex.

Therefore, servicing and conditional commitments must be treated as first-class needs, not afterthoughts. Without full context of the requests, travel retailers risk returning responses that will very likely not fit the traveler’s need.

Addressing this requires a more balanced exchange of value between AI assistants and travel providers. The more AI assistants are able to share user information, context, and intent, the more value merchants can provide in return through rich, high‑quality data. Only then will both AI assistants and travel retailers be able to effectively meet traveler needs through truly personalized offers. It is no longer only about segmentation and personalization, but about continuous intent understanding, contextual decision-making, and real-time relevance. 

While emerging standards such as UCP represent an important step, they remain largely retail-native and are  still in the process of being adapted to handle the full complexity of travel.

Closing this gap will require deep, ongoing collaboration with the travel industry and technology partners to ensure protocols are robust, context-aware, and capable of supporting end-to-end travel experiences.

Elsewhere, the travel ecosystem must evolve, being deliberate about where agentic approaches genuinely add value, focusing on areas of real complexity where proprietary data, rights, or domain expertise matter. It requires designing agents that can be invoked through low-level protocols such as A2A, both internally and, where it makes strategic sense, externally.

The industry must prepare for a world of orchestrated, multi-agent systems, where coordination across an emerging “agent mesh” becomes critical. This includes upgrading back-end and existing APIs to be AI-friendly.

The future is not “all chat.” It will involve both human-to-AI and AI-to-AI communication, where high-quality, well-governed data becomes the foundation, because humans respond to emotion, while agents respond to facts.

Agentic AI also introduces new layers of security, compliance, and operational risk, demanding stronger observability, explainability, and control mechanisms. Permissions, auditing, transparency, rate limiting, and policy enforcement must be built in from day one.

Even then, enabling true travel retailing remains a harder challenge: while commerce protocols promise seamless, multi-party transactions through standardized schemas, the inherent complexity of travel, spanning inventory, pricing, fulfillment, and regulation, makes this far less straightforward.

Protocols fit for the travel industry will have to be able to cope with travel complexity, standards and scale. We have operated online at global scale for decades, processing three billion flight searches a day across more than 400 airlines and two million hotel properties.

At Amadeus, we are transforming travel with AI. We are the embedded and neutral execution layer for travel, with foundations built at global scale, the power of our integrated and deeply connected business logic, and our status as a trusted system of record in the travel industry. We are uniquely positioned to orchestrate an AI-driven travel ecosystem, connecting suppliers, sellers and AI assistants to trusted, dynamic travel data at scale, in a neutral, secure and responsible way.

These pillars enable us to collaborate, developing AI-driven capabilities consistently alongside our partner airlines, airports, hotels, and travel sellers, as well as the wider travel ecosystem and native AI players themselves.

Our scale and deep travel expertise position us to help design agentic protocols that truly work for travel. We’re already collaborating with key technology players, and we’re ready to partner across the industry to evolve AI standards that are secure, responsible, scalable, and fit for real-world travel complexity.

Our technology is deeply integrated across the travel industry, connecting systems and workflows developed over decades. It combines industrial-grade reliability, operational resilience, and data-driven insight that enable us to deploy AI in real-world production environments globally.

Together, these pillars reinforce Amadeus' role as the trusted technology partner enabling innovation at scale. AI will augment and reinforce the Amadeus travel platform.