How can airlines get started with agentic AI?
There’s a lot of hype around agentic AI but in this piece, I’d like to look at how airlines can move beyond that and start implementing it.
Mycolleague Meg O’Keefe, EVP Airlines Americas, Amadeus, recently addressed a room of airline leaders at IATA’s AGM in Rio de Janeiro, during a panel session entitled: “Is AI Living Up to its Promises for Airlines?”
It’s an interesting question. In my view, the technology will deliver on its promises for some airlines, while falling short for others. What will determine success or failure from an airline’s investment in Artificial Intelligence (AI)? It will be defined by how each carrier approaches implementation alongside its broader technology strategy.
In our recent whitepaper ‘Airlines in the agentic age’ we combined insights from pioneering airlines, teams at Amadeus and a range of industry experts including those from Microsoft, to propose five practical recommendations for airlines getting started with AI.
Key takeaways:
- Focus on where agentic AI can have the greatest impact
- Prepare your data
- Understand how AI interacts with other technology
- Build in governance by design
- Start change management from day one
Let’s look at each in turn.
1) Focus on where agentic AI can have the greatest impact
In the immediate term airlines need to be selective, carefully focusing on a handful of workflows where AI is expected to deliver maximum return on investment. An obvious starting point is call center automation, where AI voice agents can now undertake predictable interactions like rebookings – reducing wait times and freeing human teams for more complex interactions. In other human-assisted examples we’ve seen AI agent productivity tools in the back office reduce customer handling times by 76% on average in the travel industry.
In our report, Microsoft’s Matt Walsh said it best: “Today, companies are excited about how many AI agents they’ve spun up—but very few have scaled. My advice is to focus—one agent that’s scaled and delivers a tangible return is more valuable than 100 agents in pilot.”
2) Prepare your data
Like all data analysis technology, the value gained from agentic AI will be proportional to the quality of the data it can access. For an airline, this means a key focus must be ensuring that specific data and business logic available within its own system is ready for AI agents.
Specifically, this means event data generated by an airline’s operations needs to be captured and made available in a data lake to power real-time inference. It also requires APIs to be re-engineered for AI agents. Finally, the language (or semantics) we use in aviation is very specific and fine-tuning generic AI models to understand our industry’s language is essential.
At Amadeus, we are supporting airlines in this area with our data mesh, which normalizes and governs 30 Petabytes of high-quality travel data. Through Amadeus Nevio, our portfolio of modular retailing solutions, our customers benefit from an event-driven platform that surfaces real-time data from across their operations that can be leveraged by AI agents. We have also developed a semantic layer that enhances AI models for use in complex operational scenarios at the airport.
3) Understand how technology interacts
Agentic AI will unlock more relevant customer experiences and servicing opportunities for airlines. However, it’s important to assess whether your airline’s underlying systems can support an AI use case before investing.
While airlines can benefit from AI irrespective of the underlying systems they choose to run their business, it’s important to identify system limitations from the outset. For example, attempting more granular contextualization with AI will deliver maximum value if the airline has first progressed with the move to Offer and Order standards. The airline’s products need to be cataloged for AI to personalize at a granular level.
4) Build in governance by design
Unlike prior generations of enterprise IT, AI agents are probabilistic and not deterministic. While this makes it possible for them to problem solve through reasoning, helping airlines accelerate innovation and bring new capabilities to market faster, it also means their outputs may vary depending on the context and inputs provided, and may be inaccurate or inconsistent. The challenge for airlines is therefore not only to harness the benefits of AI, but also to implement appropriate safeguards to ensure its responsible and compliant use, which is where governance comes in.
Governance cannot be an afterthought.
It’s important to consider technical, ethical, legal, and process measures for AI governance from the outset, including clear accountability, data governance, transparency and appropriate human oversight. Striking the right balance between tech-enabled governance that draws on deterministic, rule-based systems as guardrails for AI agents, and human-in-the-loop oversight for each specific use case is critical for responsible adoption and risk mitigation.
5) Start change management from day one
Introducing AI is likely to change the nature of multiple job roles. Without clear communication that explains the change, why it’s happening, and how it can benefit employees, there’s a risk of low adoption.
While AI is poised to automate transactional tasks, humans will continue to play a fundamental role. Airlines have an opportunity to redeploy human experts to new positions focused on creative thought, AI orchestration, and customer service. When the basics are automated, we can focus on more complex scenarios that can significantly improve the traveler experience or were previously out of reach.
To conclude, Agentic AI isn’t just another system migration project. Embedding artificial intelligence represents a fundamentally different approach with the potential to improve the traveler experience, unlock new revenue opportunities and improve efficiency through automation.
Yet accessing these benefits will require airlines to take both a systematic and strategic approach to AI. Simply adding a Model Context Protocol (MCP) wrapper to an API and launching AI agents is unlikely to deliver the desired return. Agentic AI should form part of an airline’s broader strategy, with clear planning, objectives and preparation being the key to long-term success.
With our foundation as a system of record, our deep integration across the industry, and our ability to run at production scale, supported by strong governance through our AI Office and strategic partnerships, Amadeus is helping the industry embrace AI with integrity and trust.
If you’re interested in understanding how airlines are beginning to generate ROI from AI then I encourage you to download our latest report.
TO TOP