What Lessons Has Amadeus Learned from Real AI Deployments?

Key takeaways

• Successful AI deployment in travel starts with a clearly defined customer or business problem. Use cases need measurable value, trusted data and integration with real operational workflows to move from pilot to production.

• AI readiness depends on governance, human oversight and adoption as much as technology. Privacy, security, compliance, monitoring, accountability and user trust must be built in from the outset.

• Agentic AI is moving travel from conversational search towards action. Reliable booking, servicing and disruption management require AI assistants to connect with trusted availability, pricing, loyalty, payment and fulfilment systems.

• Amadeus is applying and scaling AI across the travel ecosystem. Through the Amadeus AI Delivery Hub, Amadeus AI Foundry, partnerships with AI companies and solutions including Amadeus Max and SkyLink, Amadeus is turning AI experimentation into practical travel applications.

Last week I attended the Skift Data + AI Summit Europe in London. One message came through clearly: the travel industry is ready to move beyond predictions about AI and focus on what it takes to deploy it successfully.

The central lesson is that successful AI deployment is not simply about choosing the right technology. It requires a clearly defined outcome, production-ready foundations and integration with the systems that turn intelligence into reliable action.

This is a space where Amadeus has been rapidly developing expertise and can offer insight. Here we look at some practical perspectives on what it takes to move AI from pilot to production across the ecosystem, and some of the pitfalls to avoid.

Scaling real-world AI deployments

Amadeus’ Tailored Horizons research found that 74% of U.S. travellers have used AI during trip planning. This indicates that the conversation is shifting from whether travellers will adopt AI to how the travel industry can engage with them in the channels they already use.

For us, this means engaging across a complex ecosystem of global and regional AI players and connecting emerging capabilities with trusted travel intelligence, including through our collaborations with Anthropic, Google, Microsoft and others.

New entrants such as Muse and Instinct also illustrate the speed of innovation and the wider shift from conversational search towards agentic systems that can take action. Following both established and emerging players helps us focus on the trends that will meaningfully shape travel.

Our work with AI-native companies has reinforced a key lesson: the strongest use cases start with a clear business or traveller problem, then apply the right technology. AI is not the story in itself. Outcomes are.

Why do some AI use cases fail?

Not every promising AI experiment is ready to scale. The difference often lies in whether the right foundations have been considered from the outset. Many start too broadly, lack a clear workflow, or are expected to compensate for fragmented systems, weak data, or unclear ownership.

AI is also developing within an evolving regulatory landscape shaped by the European Union AI Act, GDPR and global consumer protection law.

Use cases are particularly difficult to scale when governance, context and integration requirements are not addressed early, leaving promising concepts without the foundations needed for production. Legal, privacy, compliance, governance and human oversight should therefore be embedded from the start, rather than added as checks at the end.

The customer perspective should come first. Before deciding what to build, teams need to understand the problem they are solving, who experiences it and what meaningful outcome would demonstrate value.

Successful use cases start with a focused customer problem and a clear measure of value, such as conversion, handling time, adoption, error reduction or customer experience. Scaling them also requires trusted data, integrated workflows, governance, human oversight and reusable foundations from the outset.

This is why we created the Amadeus AI Delivery Hub: to bring together business, product and AI expertise, industrialise successful use cases and build capabilities that can be reused across the travel ecosystem. Within it, the Amadeus AI Foundry turns common AI challenges into reusable capabilities, helping teams move faster while maintaining consistency, governance and control. The objective is not simply to launch more experiments, but to scale what works without repeatedly solving the same underlying challenges.

This is something my colleagues Gaëlle Bristiel and Maryline Brely touch on in their articles on moving AI from experimentation to execution and the Amadeus AI Foundry.

Why readiness, governance, and adoption matter

Moreover, AI readiness is about more than choosing the right model. In production, questions of ownership, monitoring, permissions, escalation, accountability, cost, and user trust can be just as important.

Trustworthy AI needs to be controllable, transparent, and governable, with clear boundaries around data, decisions, and human oversight. That means embedding privacy, security, compliance, and responsible AI principles from the start, alongside robust evaluation and observability to monitor non-deterministic systems in production.

Readiness also means ensuring the wider travel ecosystem can support AI-driven interactions reliably and sustainably. Finally, adoption is critical. AI only creates value when people trust it, understand it and can use it effectively in their everyday work.

Moving from search to action

AI assistants are creating a new front door to travel, but turning intent into action requires much more than adding a conversational interface. The assistant needs to connect to trusted availability, pricing, loyalty, payment, servicing, and fulfilment capabilities, with the underlying systems able to share context and expose the right state, constraints and business logic.

This changes the technical, operational and economic model of travel.

Hotel shopping illustrates the shift. Traditional APIs were built around structured search and selection, while AI can start with an outcome-based request. As AI agents make search more open-ended, travel companies also need to manage the resulting query volumes and their impact on infrastructure costs and look-to-book economics.

Amadeus Max for Travelers is the latest traveller-facing solution we have introduced. It brings agentic AI into airline customer service, providing conversational, real-time support across web, mobile and call centre touchpoints. Its first available use case is voluntary rebooking, with specialised AI agents interpreting the traveller’s request while existing airline and Amadeus systems continue to enforce business rules, data controls and transaction logic. The solution also supports a handoff to a human service centre agent when needed, keeping people at the centre of the experience.

Alongside Amadeus Max for Travel Sellers and SkyLink, it demonstrates how agentic AI is already being applied across several real travel workflows.

What advice can Amadeus offer to speed AI implementation?

Start with the customer problem. Choose a focused, meaningful challenge and define the outcome that would demonstrate value.

Design for production from the outset. Assess data, integration, governance, monitoring, permissions and human handoffs before committing to scale.

Build for the ecosystem. Connect AI to trusted travel systems, business logic and operational processes so that it can support reliable booking, servicing and disruption management.

Moreover, because travel is deeply interconnected, an ecosystem approach is essential. Amadeus aims to play a central role at this intersection, connecting travel players and AI companies through trusted intelligence, industry integration and responsible AI.

The opportunity is to solve real problems, build on trusted foundations and collaborate across the ecosystem so AI delivers better, more connected experiences for travelers.

Benjamin Cany – Vice President Product & Strategy, AI Delivery Hub – Amadeus 

Source: View the original article at Amadeus.