91% of hotel chains use AI. 13% can show a return, and most still type guest preferences by hand

A new reading of this year's h2c global study puts a number on something most hoteliers already suspected. Adoption is nearly universal. Payback is not.

The analysis, published by the hotel data platform dailypoint on 22 October alongside a free AI readiness assessment, reports that 91% of participating hotel chains already use AI technologies, while only 13% have achieved a measurable return on their AI initiatives.

The gap between those two figures is the story.

The numbers that explain it

Three findings from the same study point at why the return is missing, and none of them is about the AI.

69% of hotels using AI-based solutions still manage or enrich guest preferences manually. Only 19% use AI to extract preferences from guest interactions.

Just 42% of the hotels surveyed have a central guest profile database.

Read together, that describes an industry buying AI tools and then feeding them by hand from systems that do not share a guest record. More than half the respondents have no single place where a guest's history lives, which means a personalisation tool is working from whichever fragment it happens to be connected to.

"If the first challenge faced by AI in hospitality was about adopting the technology, the next one will be about whether hotels have the data foundation to create actual value from it," said Michael Toedt, chief executive and founder of dailypoint. "AI is only as smart as hotel data available, and it cannot resolve fragmented guest profiles or inconsistent information on its own."

The readiness check

dailypoint has built a free, anonymous online assessment off the back of the analysis. Eighteen questions across six areas, covering data foundations, governance, privacy, security, use cases, and business culture and marketing. It takes about five minutes and returns a score with recommendations.

It is a lead generation tool, and it is reasonable to say so plainly. dailypoint sells a central data platform for CRM and loyalty, and the diagnosis it has drawn from the study is that hotels need a central data platform. That does not make the finding wrong, but it does mean the recommendations at the end will point somewhere.

The assessment is worth five minutes anyway, mostly because the six headings are a decent checklist whether or not you ever speak to the vendor. Governance and privacy in particular are the sections hotels skip when a department buys an AI tool on a card.

How this sits with what we already knew

PMS News reported in late October that the same h2c research found only 15% of hotel chains have a standardised AI data access layer such as MCP, with 52% having no plans to adopt one, while 60% of day to day AI use happens in external tools like ChatGPT rather than inside vendor systems.

Those figures and these ones describe the same problem from two directions. Staff are using AI. The AI is not connected to the systems of record. So the work gets done by copying data out and pasting answers back in, and nothing shows up as a return because nothing was automated, only assisted.

What to check before you count on any of this

The study reports percentages of "participating hotel chains" without a published base number, so a single-digit swing in the sample moves several of these figures. The analysis is dailypoint's own, not h2c's.

And measurable ROI is a high bar that depends on whether anyone set a baseline. A hotel that cannot show a return may simply never have measured the before.

If you own the tech stack at a group, the useful action is narrower than a readiness score. Pick one AI tool already in use, find out which system it reads the guest from, and check whether that system holds the whole guest or a slice of them. dailypoint is competing as a semi-finalist at the HTNG and AHLA TechOvation Awards at the Hospitality Show in Miami.