Airlines say they are ready to let AI rebook you when your flight falls apart. Most still think you will wait politely in their queue while that happens. That is the bit that will not last.
At Grounded 2026 in Reykjavík, leaders from more than 35 airlines covering over 40% of global passenger volumes were asked: if the technology worked tomorrow, how much decision-making would you hand to AI during disruption? Matthew Walker’s write-up on PhocusWire has the numbers. Seventy-four percent were prepared to let AI take action of some kind. Only 26% wanted it stuck on recommendations. Full autonomy for most passengers? Just 7%.
So the industry is not anti-AI. It wants AI that acts, but inside fences. Fair enough. The problem is who those fences are built for.
The gap passengers actually feel
Ask where the biggest gap sits between what a disrupted passenger expects and what an airline can deliver, and the answer is not “I wish a human would hug me.” Forty-one percent chose speed of information. Twenty-seven percent chose personal relevance. Twenty-five percent chose quality of options. Human contact when it matters came last, at 8%.
That matches what every delayed traveller already knows. Flight trackers and WhatsApp groups often tell you the bad news before the airline does. Once AI can understand a problem in seconds, waiting hours for a rebooking starts to look like a choice, not a constraint.
Walker’s “1% problem” lands hard. During Thanksgiving 2025 in the US, 31 million passengers flew with a 99% completion rate. The remaining 1% was still 310,000 people. At aviation scale, edge cases are cities. An AI that gets 99% of rebookings right and quietly breaks the other 1% is a brand crisis with a passenger name on every row.
Ready for AI. Not ready for your AI.
Here is where I push back. The same summit asked about a future where a passenger’s own AI agent handles rebooking, choices and compensation. Seventy-eight percent said that would take years. Fifteen percent thought it was near-term. Seven percent thought it was imminent.
They may be right on the timeline. They are wrong if they think the current model survives that shift unchanged.
Today’s disruption playbook assumes three things: the passenger comes to the airline, the airline sets the menu of options, and the airline defines what “good recovery” looks like. A personal AI agent challenges all three. It will compare your rights, connections, hotel voucher and compensation claim against the airline’s offer, then negotiate. That is not a nicer chatbot. It is a change in who holds the leverage.
Airlines are busy deciding how much autonomy to give their systems. The harder question is what happens when the passenger’s system shows up with a mandate and a memory of every refund rule the carrier would rather keep fuzzy.
What is actually blocking them
When Grounded asked what is stopping AI, regulation barely featured at 4%. The business case was 10%. The real blockers were systems and data (41%), organisational readiness (27%) and trust and governance (18%). In other words, 86% pointed at readiness, not law or money.
That lines up with what British Airways-parent IAG’s chief AI scientist, Ben Dias, said at Skift’s Data + AI Summit in London. PhocusWire reported his three-question test for automation: how critical is the decision and what does failure cost; what is the blast radius if it cascades; and what is the fallback when AI fails, because it will.
Those are the right questions. Most airline marketing teams are not asking them yet, because disruption still sits in ops. That split is getting expensive. The customer who waits four hours for a rebooking does not blame “irregular operations”. They blame the brand.
Why marketers should care
If you sell seats, hotel nights or tours off airline reliability, this is a trust story. It will show up in reviews, social noise and, eventually, in how AI assistants answer “is this airline good when things go wrong?”
I have already argued that the tiny share of traffic AI sends hotels is the wrong number to watch. The same logic applies here. AI will shape the recovery narrative long before it books the replacement flight. And if your product data is a mess, AI cannot save you any more than it can invent a comfortable seat the airline never defined.
United’s Connection Saver at Denver, cited at the Skift summit, reportedly saved 54,000 passenger connections. The brands that win will pair that kind of system with a recovery story passengers can see in the first ten minutes, not the first ten hours.
What I would do this quarter
- Map every disruption decision you are willing to automate, and name the failure mode. Dias’s three questions are a good start.
- Treat passenger-facing status as a product, not a by-product of ops. Speed of information won the survey for a reason.
- Assume a passenger AI agent is coming. Design offers, vouchers and compensation rules so a machine can parse them without a phone queue.
- Connect commercial and ops data so the rebooking AI knows value, accessibility needs and the onward trip, not just the next open seat.
- Publish the playbook in plain language. When AI assistants answer “what happens if my flight is cancelled?”, you want your own policy in the source list.
If you want help making travel AI and search work together for your brand, my travel AI and GEO services page sets out how I work with airlines, hotels and operators.
The debate
Airlines are preparing to let AI act during chaos. Good. The industry that still assumes the passenger will wait in the airline’s channel is preparing for the wrong future.
So when your flight breaks, should the airline’s AI rebook you first, or should yours negotiate on your behalf?
Follow the newsletter: Dan Taylor on LinkedIn