Will the revolution be an evolution? Part 2 of 5

Airlines Are Automating Where It's Safe, Then Where It Matters

Airline leaders agree on where AI automation should begin. But the decisions with the greatest consequences for passengers are also the ones they're least ready to hand over.

In Part 1, airline leaders told us they're ready to let AI act, but their airlines aren't ready to receive it. So where do they start?

At Grounded 2026, we asked tables of airline leaders which areas of disruption management should be automated urgently, which were realistic, and which would prove controversial.

The results suggest a clear path: automate the response to disruption first, then tackle the decisions that determine its consequences.


Topic

Urgent

Realistic

Controversial

Disruption care (hotels, meals, vouchers)

57%

40%

3%

Compensation

24%

43%

32%

Passenger prioritisation

31%

28%

41%

Partner rebooking

21%

38%

42%

Scenario planning (network-wide recovery)

33%

18%

48%

Workshop votes from tables of airline leaders at Grounded 2026. Each row shows the share of that topic's votes, rounded. These are votes, not airlines.

Automating where it's safe

Disruption care was the clear favourite. 57% of votes identified it as urgent, 40% as realistic, and just 3% considered it controversial.

And it's easy to see why.

Automating the response to disruption (letting AI arrange hotels, meals, vouchers and transport) is relatively (yes, relatively) low risk. The rules are generally understood, the processes repeatable, and the benefits immediate. The optics are great, too. Passengers see an airline responding rather than leaving them to figure things out themselves.

One workshop table captured the balance:

"Automation can initiate first-step changes, but a human retains ownership."

Starting here makes sense. It delivers value, builds trust and creates a foundation for more complex automation.

But it's still largely automating what happens after a disruption decision has been made.

Then where it matters most

At the other end sits network-wide recovery. Just 18% of votes considered scenario planning realistic for automation, while 48% called it controversial. Passenger prioritisation, deciding who gets limited seats and recovery options, was similarly contested.

These are fundamentally different decisions.

Automating passenger care means responding to a situation. Automating network recovery means influencing the situation itself. Which flights should be cancelled? Which aircraft should be repositioned? Who gets the remaining seats?

Consider cancelling a flight to protect the wider schedule. Operationally, it might make sense. But what if that flight carries hundreds of connecting passengers who cannot travel until tomorrow?

AI might arrange their hotels and vouchers in minutes. Impressive. But could a better decision upstream have avoided some of those costs and disruptions altogether?

Faster vouchers after a poor operational decision are still a poor passenger outcome.

The bigger opportunity may be reducing how many passengers need recovering in the first place.

Why airlines aren't there yet

It's not necessarily that airlines doubt AI's ability to compare recovery options. The difficulty is deciding what to optimise for.

Protecting the schedule might increase passenger disruption. Minimising costs might extend recovery times. Prioritising one group inevitably affects another. And who owns the decision when AI recommends cancelling a flight? Operations? Customer Experience? The person approving the recommendation?

These are organisational questions as much as technological ones.

It's understandable that airlines begin with established rules and relatively contained consequences. But staying there limits what AI can achieve.

AI doesn't have to make the decision to improve it

The Grounded workshops didn't suggest handing network recovery decisions to AI overnight.

Instead, leaders described a practical middle ground:

"Automatically estimate and compare recovery option costs… but leave the final decision with a person."

Another table suggested:

"Keep strategic decisions with people, but use AI to vet those strategies."

AI doesn't need to decide which flight to cancel to improve that decision.

It can compare scenarios, calculate passenger impact, identify missed connections and estimate recovery costs before a human makes the call.

Each useful recommendation builds confidence. Over time, automation can evolve from supporting decisions to executing certain actions within agreed guardrails.

From safe to significant

Across the Grounded workshops, we mapped 17 topics according to urgency and controversy.


The pattern was revealing. Passenger care and communication attracted broad agreement. Network recovery and passenger prioritisation were urgent but contested. Other capabilities, including predictive risk and AI-assisted recovery decisions, remained controversial without attracting the same urgency.

The map uses the median urgent and controversial vote shares across 17 topics to define its quadrants. It indicates direction, not a precise ranking.

Airlines have a sensible starting point. But the real test is whether those early successes become a stepping stone towards more consequential decisions, rather than the destination.

Because the greatest opportunity isn't simply automating how airlines respond when things go wrong.

It's improving the decisions that determine how badly things go wrong in the first place.

Next in the series

In Part 3, we examine why AI will optimise whatever airlines measure and why measuring the flight rather than the passenger could produce very different outcomes.

Data: Workshop votes from leaders representing 35+ airlines at Grounded 2026. Results are directional, not a market-wide survey.
Part of a series based on Matt Walker's article "Will the revolution be an evolution?", first published in PhocusWire. Read the full article.