A review of what AI can and cannot solve.
Flight scheduling is usually discussed as an optimization problem: fit more flying into a fixed fleet, improve aircraft utilization and match capacity more closely to demand. But scheduling also has a resilience function. When weather, strikes, maintenance requirements, airport constraints or sudden demand changes disrupt the original plan, the quality of the schedule influences how much flexibility an airline has left.
Artificial intelligence can help planners explore more options, process more variables and test alternative schedules faster than traditional manual approaches. That can strengthen resilience because airlines have more opportunity to understand how a network might respond before disruption occurs. It does not, however, remove the physical and regulatory constraints that determine what is actually possible. AI can improve the speed and breadth of decision support. It cannot create a spare aircraft, open a closed airport or make an unavailable crew suddenly legal to operate.
Resilient schedules need room to absorb disruption
A highly utilized fleet can produce attractive economics when operations run to plan. The same schedule can become fragile if there is little room to recover from delay. A late inbound aircraft may affect several later sectors, while a maintenance requirement or unavailable crew can force changes across multiple routes.
This is why resilience should be considered when the schedule is designed rather than only when irregular operations begin. Planners need to balance aircraft productivity with turnaround assumptions, maintenance windows, crew availability, airport slots and the likelihood of delay at different points in the network. A schedule that looks efficient under one set of assumptions may prove much less efficient once disruption is introduced.
AI-powered tools can help by evaluating larger numbers of combinations and running scenarios that would be cumbersome to test manually. This gives planning teams more opportunity to identify where a schedule is especially sensitive and where a small change could create more recovery capacity.
Scenario planning is where AI can add resilience
The strongest resilience case for AI is not that an algorithm can predict every disruption. It is that planners can test more possible responses before they need them. Traditional scheduling processes can require extensive manual adjustment, limiting the time available to model alternative operating conditions.
KPMG in Ireland’s 2022 work on AI flight scheduling found that scheduling teams can spend weeks or months adjusting draft schedules and may have insufficient time to test alternatives. The research argues that faster scheduling tools can free analysts to do more scenario planning and war-gaming, improving their ability to examine different network designs and responses to operating shocks.
For resilience planning, useful scenarios might include:
- the temporary loss of an airport, route or base
- reduced aircraft availability because of maintenance
- changes in demand that require capacity to move quickly
- crew or slot constraints affecting selected parts of the network
Running these scenarios does not tell management exactly what will happen. It can show which parts of the schedule have alternatives and which dependencies could create disproportionate disruption.
AI still operates inside real-world constraints
The limitations matter as much as the potential. Flight schedules are built around rules and resources that cannot simply be optimized away. Landing and departure slots have conditions attached to them. Aircraft need to be in the right place for maintenance and inspection. Crew availability is constrained by qualifications, working-time rules and positioning. Airports have operating limits and connections may need to be protected.
The Aviation 2030 material identifies these variables as part of the complexity scheduling analysts already manage. It also notes that full-service carriers can face additional constraints, such as maintaining banked connections or particular arrival and departure windows.
That is an important boundary for AI. A system may find a mathematically attractive schedule, but the schedule still has to be operationally workable. Resilience therefore depends on the quality of the constraints, assumptions and data supplied to the model, as well as the judgment of the people interpreting its output.
Faster rescheduling can improve the response to irregular operations
AI may also have value once disruption has started. If scheduling tools can generate viable alternatives quickly, operations teams may be able to compare recovery options without manually rebuilding large sections of the plan. The KPMG research specifically identifies irregular operations such as bad weather, strikes and short-term schedule changes as an area where dynamic rescheduling could help reduce delays and unexpected costs.
This is where Aviation 2030 insights provide useful context for the wider role of digitalization in aviation. The series places AI flight scheduling within a broader discussion about how technology is changing airline operations, while also recognizing the complexity of adoption in a regulated and operationally constrained industry.
The distinction is important. Faster rescheduling can help an airline choose among feasible recovery options. It cannot remove the underlying disruption. If aircraft, gates, crews or airspace are unavailable, the technology is working with a smaller set of choices. The value comes from identifying those choices quickly and understanding their consequences across the rest of the network.
Resilience still depends on people, process and operating slack
AI can make flight scheduling more responsive, but resilience cannot be delegated to an algorithm. Airlines still need clear decision rights, reliable operational data, experienced planners and contingency plans that account for the realities of the network. They also need to decide how much efficiency they are prepared to trade for recovery capacity.
A schedule with no operational slack can remain vulnerable even if it was produced by sophisticated technology. Equally, a flexible schedule is of limited use if disruption decisions move slowly between planning, operations, crew, maintenance and commercial teams. The technology works best as part of a wider operating model that can act on its recommendations.
The strategic opportunity is therefore narrower, and more useful, than the idea of AI solving airline disruption. AI can help teams test more schedules, identify vulnerabilities and respond faster when assumptions change. It cannot eliminate the constraints that make aviation difficult to operate. Resilient scheduling comes from combining stronger analytical tools with realistic assumptions, human judgment and enough flexibility in the network to make alternative decisions possible.