Market Intelligence
Market Intelligence
98% of hoteliers already use AI in some form. The gap is not adoption. It is the difference between plugging in tools and building AI that actually understands how a property operates.
The Central Idea
The hotels that will win with AI are not the ones who found the best vendor. They are the ones who understand their own operations well enough to know what to automate, what to protect, and how to make the technology genuinely theirs.
The survey says 98% of hoteliers use AI and 92% feel optimistic about it. But adoption is wide and shallow. Someone using ChatGPT to write marketing copy counts. That is real, and it is a long way from AI that understands a specific property and helps run it.
What Hoteliers Actually Want
The tech world assumes the goal is removing humans from as many steps as possible. In real conversations with hoteliers, that framing rarely holds.
The instinct from technology is that full automation is the end state, and every human step is friction to remove. Madeline kept bracing for conversations to go this way. More often than not, they did not.
The most excited hotels are the ones whose GM says the front desk spends two hours a day on operational tasks that have nothing to do with guests, and wants that time given back so the team can focus on relationships and connection.
Guest-Facing AI
Hoteliers happily automate back-office pricing. Hesitation appears the moment AI touches the reception desk and the guest relationship. The caution is rational.
It might cost some revenue on a slow Tuesday. That is not ideal, but the revenue is recoverable. This is why revenue management was automated first.
A guest drove four hours for their anniversary weekend. A botched complaint can end that relationship for good, and the review ends up on TripAdvisor. Hoteliers understand this intuitively.
The efficiency comes from removing the friction around decision-making, not removing the decision-maker.Madeline Bushbeck
The Copilot Model
Hospitality is a human-centric industry by definition. The internal framing is a copilot: AI surfaces the right information, drafts the right message, and flags the right moment. A human decides.
The system learns from what staff accept and what they reject. Once acceptance rates are consistently high, around 95%, it becomes reasonable to ask whether staff are comfortable switching on automation. The foundational, compounding trust has to be built first.
The Semantic Layer
Every hotel operates differently. The semantic layer is a contextual knowledge layer that AI agents sit on top of, so they can reason the way anyone at the hotel would reason.
AI Agents
Make suggestions on pricing, communications, and how to accommodate a guest.
Semantic Layer
Room types, services, policies, guest history, and the institutional knowledge that lives in staff heads. It is also a learning loop: it updates as real events contradict the initial draft.
Connected Property Data
Reservations, profiles, operations, and communications living in one place.
Front desk staff knew that all rooms ending in 10 are suboptimal and should be filled last. That fact exists nowhere in the PMS. An agent recommending a room needs that knowledge to make a good decision. The semantic layer is where it now lives.
A fixed set of definitions the AI is supposed to follow. It does not change as the hotel changes. Useful, but inert.
A learning loop. Based on the actions front desk and revenue managers take, it keeps learning and improving. An evolving model that follows the staff team rather than a document they have to maintain.
Start With Your Data
If your systems do not talk to each other, your AI will not either. Going from scattered experiments to a real strategy starts with connecting the data.
Historically, hotel groups built data lakes outside operations, purely for reporting. Reports are not actionable. Mews moved the storage logic inside the PMS, the heart of action, where the operation actually runs. Because Mews is the PMS, the point of sale, and the payment solution, data points connect that could not before. A restaurant transaction paid with Apple Pay tracks back to the guest profile.
Using AI Badly
You just get bad outcomes faster. Without good data and good information underneath, throwing AI on top cannot produce good results.
Five different AI vendors with no shared context. None of the systems know what the others know, so the combined result is poor.
Misrepresenting what a property can offer creates an immediate trust problem with the guest and reflects badly on the hotel.
A tip suggested a wake-up call. Most properties offer wake-up calls as an undocumented complimentary service, but this hotel ran with no front desk, so there was nobody to make the call. The tip was neither actionable nor relevant. The fix is not better AI; it is AI grounded in the actual reality of the property, which is exactly what the semantic layer stores. Hospitality is an industry of edge cases, so Smart Tasks also had to be configurable per property (one hotel only wanted action items for VIP guests).
Saying AI-powered effectively means nothing. Here is what it does for your front desk actually means something.Madeline Bushbeck
Real AI vs Marketing
Everything is marketed as AI-powered. To cut through it, a hotelier can ask a single question: does the AI know anything specific about my property?
A product layered on the outside that knows nothing about your operation until you do the work to teach it. Not very effective.
The institutional knowledge in staff heads, available without a lot of manual setup. AI as an integration, not just a product.
Skip the demo sizzle. Name the concrete output.
Tie it to a business outcome you can measure: revenue, or guest satisfaction that drives return guests.
An AI that answers a guest incorrectly, or kicks the question back to the front desk anyway, only adds noise and work for both sides.
AI is expensive. The investment must drive an outcome, not a gimmick. Anchor to the ROI metric.
Building Conviction
The model landscape changes faster than the tech industry has seen before. The answer is to stay anchored to the problem rather than the solution.
What does not change: hotels need to serve guests better, operate more efficiently, and do more with leaner teams. Build continuously toward that, and the specific infrastructure and models become an implementation detail rather than the core problem. If something did not work the first time, that is not a reason to drop it. Iterate, and revisit quickly rather than waiting a year.
The One Idea
The winners are not the ones who found the best vendor. They are the ones who know what to automate, what to protect, and how to make the technology genuinely theirs.