By Eric Léopold, Founder, Threedot 

The ‘Artificial Intelligence in Aviation’ track

As chair of the artificial intelligence (AI) track at the upcoming World Aviation Festival, I have the privilege of moderating several panels on AI in aviation. In preparing these sessions together with panellists from airlines, technology providers, and industry organisations, I gathered valuable insights into the themes that will shape our discussions on stage.

In this article, I would like to share a preview of some of the topics we’ll explore. My hope is to offer readers—and especially future participants—a fresh perspective on the role of AI in aviation, and to inspire you to join us and hear directly from the experts in Lisbon.

AI use cases in aviation

Airlines did not wait for artificial intelligence to develop algorithms and software that can solve some of the most complex challenges in the business, such as optimising revenue across a perishable seat inventory or optimising the scheduling of flights and the allocations of aircraft and crew. For decades the field of Operations Research has attracted PhDs to airlines, bringing the highest mathematical rigor to decision making where massive datasets and countless constraints collide daily.

Artificial Intelligence builds on this tradition and extends beyond optimisation. Today, in a complex environment, AI provides forecast, decision making support, pattern recognition, and more recently natural language capabilities, from voice generation or text summarisation. This new technology expends the scope of new use cases across the value chain.

In the customer domain, for example, airline call centres support customers with booking, changes and more services. In the short term, the language capabilities of AI can increase the productivity of the agents, by listening to conversation and executing tasks in real time while the agent remains engaged with the traveller. Eventually, once the AI is sufficiently trained, it can handle conversations at scale, in a much friendlier way than current clunky “press 1, 2 or 3” phone menus.

On the operational side, airlines’ optimisation algorithms already provide theoretical solutions under all the given constraints. However, in the case of disruptions, airlines need to make decisions in real time if a flight needs to be diverted or delayed or cancelled, due to the weather, or a technical issue or else. AI can simulate the consequences of potential actions to help with decision making and learn from similar experience and decisions.

Support functions play a critical role in enabling airline operations as scale. For example, every flight generates invoices for services from suppliers such as fuel, catering, parking or de-icing. AI can forecast costs and reconcile invoices while detecting anomalies, such as a de-icing charge when the temperature was never close to freezing point.

AI models and squads

Within airlines, the teams driving these AI solutions are typically called ‘data science’, often reporting into the Chief Digital Officer. Their foundational role is to collect raw data (on fleet, flights, passengers, weather, and more) and to transform it into reliable ‘clean data’. This data is then stored into lakes or warehouses hosted by third-party cloud platforms, for future exploitation.

The real science begins when this data is applied to solving real problems. The AI scientists develop optimisation models for specific use cases combining relevant AI methods like language models, computer vision or predictive models. Once a model is developed and tested, it can be deployed into production by the Machine Learning Operations (MLOps, as in DevOps) team, who maintains it. At this stage, digital product owners manage the lifecycle of the AI applications, collecting future requirements and ensuring valuable outcomes.

In an airline’s agile setup, work is often organised around ‘squads’. One squad may build a digital twin for ground operations, another squad may work on predictive models for aircraft or engine maintenance, and a third squad may deal with the continuous pricing of dynamic offers. A squad may bring together data scientists, an MLOps engineer, an aviation domain expert and a digital product owner, for the right mix of technical and operational skills.

As airlines move up the digital maturity curve, these squads can evolve into cross-functional centres of excellence for AI models. Ultimately, such AI centres may blend data, processes and decision logic into agentic AI, like a digital brain that virtually runs the airline.

Questions to our panellists

With these airline use cases and AI trends in mind, I will ask questions to the panellists on stage, such as:

  • What difference is AI making in airlines compared to Operations Research?
  • Is your main challenge in AI projects to demonstrate the benefits realisation?
  • Overall, is AI a tool that increases human productivity or a paradigm shift that replaces humans in certain tasks?
  • How reliable are your AI models in operational context where safety and accuracy cannot be compromised?
  • What is the most impactful use case for implementing AI in an airline today?
  • As airlines build AI centres of excellence, is there a risk of creating a new silo in airline organizations?
  • Is there still hype in today’s AI rhetoric, or is AI becoming a long term strategic pillar?

Of course, I look forward to hearing the questions from the audience as well.

Join us at World Aviation Festival 2025, where Eric will be chairing our AI Spotlight Sessions. 

For more from Eric Léopold see: