Publications / Report
Report for hospitality · Outlook 2027 · An independent briefing
Who holds the keys?
Twelve hotel AI trends for 2027: what AI may now decide for a hotel, what has been shown to work, and what to switch on next.
Riccardo Giacometti, Cross-PointOctober 202648 pagesFree, no registration
What it is about
In 2027, the hotels that benefit from AI will be the ones that decide deliberately what these systems may do on their behalf, and that measure what actually changes.
A hotel already knows how to grade authority. It issues guest keys, floor keys and master keys, and it knows who holds each one. The same grading fits AI, and it runs through this report.
- Author
- Riccardo Giacometti, Cross-Point
- Scope
- Twelve trends, 120 sources cited
- Evidence
- Published evidence to 9 October 2026; author commentary and the verification update to 16 October.
- Independence
- This report is free, has no sponsor, and no vendor or operator reviewed it.
Seven findings for 2027
Deployment is outpacing proof, on the outside and the inside.
- Finding 01
Attention is moving into AI answers faster than bookings are.
AI-referred visits to US travel sites more than doubled year on year by July 2026, from a base the report does not state. Booking Holdings says language models send it less than 1% of room nights.
- Finding 02
The two AI answer platforms monetise differently.
OpenAI places labelled ads, kept separate from the answer. Google tests sponsored units inside the answer. Rules learned on one do not transfer.
- Finding 03
The autonomy that is spreading is bounded.
The useful question about any system: what may it do without asking, and what happens when it cannot?
- Finding 04
Guest conversation has the broadest deployment and the thinnest evidence.
The design of the handoff to a person decides whether these systems help or harm.
- Finding 05
Returns hold up first where the starting point can be weighed or metered.
Food waste, energy and labour hours. Even there, the best-known programme published two results on different baselines, and corrected one figure after questions.
- Finding 06
Permissions are the decisive infrastructure question.
Who controls the property system, the data and the rules for agents will shape costs more than the choice of AI model.
- Finding 07
Emerging capabilities have a 2027 status.
Computer-use agents have no hotel deployment yet; training simulations are unresolved. Service robots are buyable. Humanoids are not ready.
Working pages in the report
These pages are made for working with: to classify your own systems, for the conversation with a brand or operator, and for regular review. They are part of the report, not separate files.
The key cabinetTwelve trends, three kinds of hotel, four actions: for the independent hotel, the full-service hotel and the multi-property company.
Page 10The authority ladderClassify a system by what it may do without asking, and by what happens when it cannot.
Page 16The owner’s pageFor the owner of branded or managed hotels: what to ask the brand or operator in writing, and what to settle in the contract.
Page 36The owner’s first 30 daysOne recurring task, tested in real work, within 30 days.
Page 37The quarterly checkThe watchlist as a register: one line per signal, four checks a year.
Page 40
Download the report
Free, no registration. The file is unchanged, exactly as published.
-
Who holds the keys? Twelve hotel AI trends for 2027, 48 pagesGerman version
PDF · 9.3 MB
From the same series
The second report in the Cross-Point outlook series, after “The Invisible Shortlist”, the briefing on how AI influences which hotels guests consider.
If this report raises questions for your business, you can write to Cross-Point and describe the specific challenge.