At the latest World Aviation Festival’s digital operations track, what stood out was not the release of a new standard, the implementation of a solution by an airline, or a product announcement by a vendor. It was something more fundamental, which will ensure 2025 remains a milestone.

Airline operations leaders and heads of data science were onstage in panels that I had the privilege to moderate, debating the future of AI for flight operations planning and control, particularly inside the airline’s Operational Control Center, where live decisions are made. In 2025, the conversation has moved from “what AI can do” and “is there an AI bubble?” to “where does AI help operators” and “how do we bring AI on board quickly and safely?”

Across my four panels, which I described in a previous article, the theme that emerged is: digital operations are not only about adding digital tools to optimize operations. It is about 1) rethinking and reshaping how airlines plan, decide, and recover operations, and 2) learning from this cycle, using AI as a key enabler, not a separate layer that cannot access commercial and operational data.

From optimizing plans to continuous planning

Planning is not a new territory for optimization. Airlines already use sophisticated optimizers for schedule construction, tail assignment, crew pairing, fuel efficiency, and more. There was no debate about the need for optimizing plans, as airlines may be at various levels of maturity depending on priorities and investments.

What is changing is the nature of operations planning itself. Rather than producing a single “optimal” plan, a few days ahead of the day of operations, airlines are moving toward continuous planning (and replanning). Indeed, the environment evolves continuously (weather forecast, maintenance schedules, crew availability, etc.), which requires adjusting assumptions continuously, multiple times before the day of operations eventually begins.

“Morning readiness” was discussed as a rolling process rather than as a static checkpoint.  Schedules can be dynamically built and adjusted with expected delays, curfews, and congestion already factored in, rather than assuming ideal conditions as planned.

As one panellist put it (I’m paraphrasing): “The value is not only in finding the perfect plan, but in being ready to adapt the plan when reality deviates.”

Data science teams are more and more equipped to model uncertainty, while operations teams require planning outputs that are explainable and actionable. In other words, optimization that cannot be adjusted quickly, or understood by the people running the airline ops, carry less value.

In summary, the takeaway from these discussions was: planning is becoming continuous and iterative, and AI’s role is to support this cycle and new agility, not to create another silo.

Day-of-Ops: Supporting multi-objective decision making

Planning is about optimizing readiness under operational constraints. The day of operations requires decision making, in real-time, based on multiple objectives: customer (satisfaction), commercial (revenue), financial (costs) and operational (on-time). Safety remains the non-negotiable constraint.

In the previous section we mentioned AI used for forecasting: predicting delays (building schedules that tolerate them), estimating taxi-out times (based on traffic control), modelling knock-on effects (based on potential disruptions), or identifying diversion risks (related to weather events).

Panellists debated how AI can support decisions, not just the models that made the predictions. In this context, AI is used to build what-if scenarios (what if we delay this flight?), compare trade-offs (protect loyal customers or children’s connections in priority?), and help humans strike the balance between competing objectives listed above.

As one speaker said (paraphrasing again): “AI is very good at telling us how to get ready and what might happen. The hard part, where human experience matters today, is deciding what we do about it.” The decisions are particularly challenging as a single operational constraint can break them, for example if the decision means exceeding crew duty limits, or curfew time, or availability of gates.

Given the balance of multiple objectives and the non-negotiable safety priority, human experience and judgement were not questioned, but AI’s value lies in supporting it with options, ranking (green-amber-red), simulating outcomes, and highlighting consequences or second-order effects that may be hard to visualize or anticipate in real-time.

Another panellist (airline chief operations) concluded: “We will onboard AI agents in the OCC in 2026 like we’ve onboarded human employees. We need to define clear roles and limits, and to build trust over time.” The takeaway was not autonomy (yet), but decision quality, speed, and confidence.

The-day-after: Recovery from disruptions, and learning from decisions

Several discussions addressed what happens after flight operations. In case of disruptions (delays, diversions, cancellations…), the recovery phase stabilizes the network, recovers crews and aircraft, and understands where things did not go accordingly to plans. The performance is measured in terms of shortening recovery (e.g. after a big storm) until normal operations (all passengers at destination, all crew and aircraft in place).

One panellist observed that learning opportunities remain under-exploited (I paraphrase): “We generate a lot of data during operations (weather, engines…), but we don’t always close the loop back into planning and decision-making.” I would add that teams don’t have time to analyse every disruption, and learning requires assessing what would have been a better decision, which is not easy.

The insight on post-operations from the panel was that AI will help digital operations maturity in terms of recovery scenarios and learning from decisions, in addition to planning for excellent on-time performance on the good days.

Our conclusion: What previous transformations taught us

The panel discussions ended on a word of wisdom or caution, from leaders who have been through other transformations. Who leads this AI transformation and how will we make it a success?

Will operations teams take ownership of AI-supported decision-making in the OCC? What role for the IT and data science teams? How will commercial and financial teams get involved to maximize benefits?

One panellist reminded the audience (I cite from memory): “AI implementation is 30% technology and 70% change management.” Which echoes a 10-20-70 rule by BCG that estimates that the algorithms represent 10% of the transformation, data 20% and the rest, 70%, is about people and processes.

Whether 2026 becomes the year when OCCs formally onboard AI “operators” remains an open question… to be discussed at the next World Aviation Festival!

For more insight from Eric, see: