78% use AI, 6% have a strategy, 22% have the data
Three figures, read in a row, tell the full story of hotel AI better than any conference: 78% of chains already deploy it, only 6% have a corporate strategy for it, and only 22% have a centralised data structure capable of supporting AI and automation tools. All three come from the same place: h2c's AI and automation study of 171 hotel chains (2025) — and buying has only accelerated since: by April 2026, Cloudbeds put chain adoption at 80%. The industry does not have an adoption problem. It has a foundations problem.
Buying runs far ahead of readiness
The purchasing enthusiasm is well documented. Canary Technologies' survey of more than 400 hotel IT decision-makers (2026 — Canary sells hotel AI, duly noted) quantifies it: 82% will expand their AI use this year, 85% will devote at least 5% of their tech budget to it, and 71% perceive a significant or transformative impact. In the franchise segment, Wyndham's owners report (325 owners, 2026) adds a telling nuance: 98% already incorporate AI in something — and 73% want to do more but feel overwhelmed and unsure where to start.
That is the portrait: almost everyone buys, almost nobody has the blueprint. And when investment arrives before strategy, the usual outcome is not disaster — it is something quieter: pilots that never scale, underused tools, and the feeling of having paid for magic that doesn't happen.
The real hole is in the data
Why doesn't the magic happen? Because AI systems feed on data, and hotel data is, in general, broken. The figures in Cloudbeds' annual report (2026, across 90 million reservations in 180 countries — a PMS vendor, also noted) are eloquent: 67% of independent hotels place managing disparate systems among their biggest operational challenges, and 4 out of 5 lose between one and two working days a week manually consolidating reports across platforms that don't talk to each other.
Add customer data quality: in mid-sized chains, between 15% and 25% of profiles are duplicated, and 20% to 30% of bookings arrive without an email or with insufficient data (Fideltour, 2026 — a CDP vendor). The classic illustration circulating in the industry sums it up: the same guest appears as "Mr. Smith" in the PMS, "John Smith" in the CRM and "J. Smith" in the loyalty system. To a human they are one person; to an automated system they are three strangers — and any AI personalising on that base will personalise wrongly three times.
The conclusion running through the year's industry analysis fits in one sentence: AI amplifies what exists, inefficiencies included.
The priority hoteliers themselves set
Strikingly, those who have been through it are clear: 70% of hoteliers put integration with the systems they already have among their top priorities when investing in AI (h2c, 2025). It is the hotel version of the advice this blog already covered regarding agents: data first, tools second.
The sensible order, distilled from all these sources: first, connect the systems that don't talk (or reduce their number); second, clean and deduplicate customer data; third, document the operations that today live in heads and paper; and only then choose the AI tools — which at that point will work first time and cost less to integrate.
My reading
Fifteen years of tour operator operations taught me that technology projects don't fail at the demo: they fail at the exceptions the demo never covered. The demo always works because it uses lab data; the day-to-day fails because it uses yours.
That is why the most useful figure in this article is not the 78% adopting but the 22% prepared. If your business is in that 22% — connected systems, clean data, documented operations — almost any AI tool will deliver, and this is an excellent moment to accelerate. If it is in the other 78%, this year's best "AI investment" probably doesn't carry the AI label: it carries the data plumbing label. It is less exciting to tell at a trade fair. It is what separates those who operationalise from those who collect pilots.
