Before AI Does It Alone, Ask the Three Questions an Airline Asks
63% of hotel companies now use AI regularly. Only 5.6% put its benefits into a budget or a P&L.
That's the new State of Hotel AI survey — 107 hotel company leaders, presented at the Destination AI summit. Adoption is done. The open question is a different one: who answers when the machine gets it wrong?
- 12.6% have a strategy, governance and execution at scale.
- 35% don't measure AI's return at all.
- 67% name change management, data or integration as the top barrier. Only 8% blame the AI model itself.
- Nobody plans to spend less on AI next year.
The technology isn't what's missing. A way to decide is.
How an airline decides
IAG — the group behind British Airways and Iberia — runs every automation idea through three questions. Its chief AI scientist laid them out in London last week:
- Criticality. If it fails, what does it cost — in revenue, safety, customers, reputation?
- Blast radius. Does the failure stay contained, or does it spread to other systems?
- Fallback. What's plan B when the AI goes down?
A proposal that fails the test isn't killed. The team is asked what safeguard it can add. And one line worth keeping: governance is easier where a human makes the final call.
The same test in a 30-room hotel
Run three everyday tasks through it:
- "Is parking included?" at 2am. A wrong answer costs little, stays in one conversation, and the front desk reads the thread in the morning. Automate it today.
- Changing a rate or availability. A mistake travels to every channel in minutes. The machine proposes; a person approves.
- A complaint or a refund. Money and reputation on the line. The machine prepares the case; a person decides and replies.
Most of a hotel's daily conversation sits in the first group. That's exactly why it's worth handing over — and why the other two keep a human in the loop, on purpose.
One more rule: a number and an owner
The survey's author put it bluntly: "adoption is no longer the problem. Accountability is."
For a small hotel, accountability fits on a sticky note. Every AI tool gets one number it should move — reply time, bookings, hours saved — and one person who checks it every month. No number, no owner: it goes.
And in Europe, the AI Act's transparency rules now require telling guests when they're talking to a machine. A clear plan B and an honest "I'm the hotel's assistant" travel together.
The takeaway: before you let any AI act alone, ask the airline's three questions — what does a mistake cost, does it spread, and who catches it. If the answers are "little, no, the team in the morning", switch it on. If not, keep a person in the loop. AI is my Wingman — and a good wingman always flies with someone watching the radar.
Diego
Hospitality Net — "Not Done Weekly — Destination AI Summit Recap!", Sloan Dean (October 5, 2026): State of Hotel AI survey by Josiah Mackenzie and Naz Aydin, 107 hotel company leaders (August–September 2026): 63.3% use AI regularly; 12.6% report strategy, governance and scaled execution; 35.2% don't measure AI ROI; 5.6% commit AI benefits to budgets or P&Ls; 67% cite change management, data quality or integration as top barriers vs 8% AI model limits; no respondent plans to spend less.
PhocusWire — "BA-parent IAG applies a 3-question test for AI automation" (October 8, 2026): Ben Dias, chief AI scientist, at the Skift Data + AI Summit in London: criticality, blast radius, fallback; a 10-principle AI policy including human oversight.
European Union — AI Act, Article 50 transparency obligations for AI systems interacting with people (applicable from August 2, 2026).
