AI for hotels: four different things under one name, and what to ask before buying any of them

What is sold today as AI for hotels

Anyone typing that search gets, all jumbled together, adverts for products that have nothing to do with one another. That is not the buyer's fault: the «AI» label is now hung on almost any hotel tool. Before asking for a demonstration it helps to separate the areas, because each tackles a different problem, is decided by a different role and is judged with different questions.

The right-hand column is there to save you time. Bedel, which publishes this guide, works in one row and a half. About the rest we speak as anyone helping you buy would: in general and without recommending anybody.

AI that decides what to charge and forecasts demand

A program watches the bookings coming in, the calendar, what happened in earlier years and what is going on nearby, and suggests how much to ask for each room each night. It also works out how many people will come, a figure later used to plan shifts and purchasing. It is decided by whoever answers for revenue: a revenue manager if there is one, the commercial director or the owner in a small hotel.

Questions for the demo:

• Which of our data does it work with, and where does it take them from? If it needs the booking system, who builds that connection and who looks after it later.
• Does it change the amounts by itself, or does it propose and a person decide? Ask to see where the limits are set.
• Does it explain the reason for each recommendation, or just hand over the figure?
• How does it behave in an odd week, with a trade fair or a large cancellation, of which it holds no history?
• How much history does it need to be useful in a hotel of our size?

AI that answers the guest: chat, WhatsApp and voice

This is what almost everybody pictures on hearing the phrase. An assistant replies on the website, on WhatsApp or on the phone to what gets asked a hundred times: what time check-in is, whether there is parking, how to get there from the airport. Some also accompany the booking or note down requests during the stay. It is decided by the front desk, the reservations department or marketing, depending on who feels the phone most.

There is a boundary here that deserves attention. Answering is one task; getting the request carried out is another. If a guest writes to the assistant that the washbasin is leaking, the chat ends well and the washbasin goes on leaking. The kinds of messaging product that exist for a hotel, and which one handles each side, are in WhatsApp for hotels.

Questions for the demo:

• Where do the answers come from: a document of ours that we can correct, or what the model knows in general?
• What does it do when it does not know? Ask to see an unanswered question and which person it appears to.
• Does it tell the guest they are talking to a machine? The EU AI Act asks that anyone talking to an AI knows it: that is AI transparency.
• Which languages does it really reply in, and who has gone over those replies?
• When someone asks for something that has to be carried out, a pillow or a repair, who does the assistant pass it to and how is it checked that it was done?
• Who can read the conversations and where are they stored?

AI in the back office and in reviews

It is the least showy of the four. In the office it reads supplier invoices, cross-checks documents and leaves ready the report that used to take an afternoon. In reputation it sorts what the published opinions say and writes a draft reply for someone to go over. The first is decided by administration; the second, by quality, marketing or management.

Questions for the demo:

• Is the reply to a review published by itself or approved by a person? A text in the wrong tone stays on show with the hotel's signature.
• Does it tell a complaint about service from one about the building? The second should end up with maintenance and not in a chart.
• When it misreads an invoice, who spots it and how is it put right?
• Which accounting program does it work with, and what happens the day you change it?

AI in operations: housekeeping and maintenance

This is the part the guest never sees and notices most. In housekeeping there are programs that order which rooms get done first, according to who is leaving and who is arriving, and balance the load among the room attendants. In maintenance AI shows up in three places: in the report of what somebody finds broken, in the calendar of recurring jobs and in watching machines through sensors. The head housekeeper, the maintenance manager or the operations director decide.

Of those three places, the report is the most modest and the first to be missed. A room attendant discovers a loose tap mid-morning; that room has an arrival early in the afternoon. Between one moment and the other there is a short while, and the usual thing is to spend it telling someone, writing it down and tracking a person down. Full hotel maintenance products are compared, criterion by criterion, in maintenance software for hotels, and what exactly a CMMS is, in CMMS.

Questions for the demo:

• Who can report a fault and how long does it take with their hands full? Have someone from housekeeping try it, not the manager.
• Does every employee need a user account and an installed application? What that involves, in maintenance app.
• How does it reach the outside technician, the lift engineer or the air-conditioning firm, who will never open your program?
• How does the front desk find out that the room can now be handed over?
• If it promises to get ahead of failures, which sensors does it rely on, and who fits and checks them?
• Which dated inspections does it bring, and do they match the ones the law requires of you? The list, in hotel regulations.

How to choose: six steps, in this order

• 1. Name the problem, not the technology. «We want AI» is not a brief. «In August the front desk cannot keep up with the phone» or «what housekeeping finds broken reaches maintenance late» are.
• 2. Find its row in the table. If the problem touches two rows, that is two purchases, and nearly always from two suppliers.
• 3. Name an owner inside the hotel. A product with nobody to look after it ends up unused, however good it is.
• 4. Take a case of your own to the demo. A real message from a guest, a real invoice, last week's fault. With examples prepared by the seller everything works.
• 5. Ask what happens when the AI gets it wrong or does not know. Who finds out, by which route, and what the guest or the employee sees in the meantime.
• 6. Count the part that falls to you. Connections with the booking system, data to load, people to train, and where guest and staff data are kept.

Start with whatever takes most hours or most patience from you today. Nobody needs all four areas, and still less all at once.

The only part Bedel does: the fault, from housekeeping to the technician

Bedel covers a short and very specific stretch, and it is not a CMMS. The hotel has a WhatsApp number and a list of its people: housekeeping, front desk, maintenance. Whoever is on that list and comes across something broken says so there in whichever way is quickest, with a few words, a photo or a voice note. Bedel's AI then does the one thing it knows how to do: understand that message and turn it into a ticket with a number and a time.

That ticket travels to the in-house technician or to the right outside supplier, and each receives it on their own WhatsApp, with nothing to install. The person who reported it, whether from housekeeping or the front desk, is told each step, so they chase nobody to learn how it is going. Whatever the hotel has marked as urgent rings the on-call phone at any hour. If the deadline runs out with no answer, the report escalates. Three times are kept for each fault: reported, accepted and closed. The whole lot exports to Excel. And once a month a reminder of the inspections that carry a date arrives over WhatsApp.

The purpose is a single one: that what the room attendant discovered in the morning is in the hands of whoever repairs it before the next arrival. How it looks in a hotel, in issue management in hotels. The same system in other businesses, in issue management over WhatsApp. And if you are comparing products for handling faults, there are two guides: fault management software and corrective maintenance software.

What Bedel's AI does not do in a hotel:

• It does not talk to guests, and a guest cannot open a ticket: what they say at the counter is written in by the front desk.
• It does not take bookings or decide what is charged for a room.
• It does not answer reviews.
• It does not connect to the PMS or to any other program.
• It keeps no equipment inventory and makes no work orders with materials, hours and costs.
• It stores no certificates or inspection records. Which products do, in mandatory inspections software.
• It does not show several hotels on one screen: each hotel is one account.

Data are stored in the European Union. If what you are looking for sits in any other row of the table, Bedel is not your product, and we would rather you read it here than find out halfway through a call.

Can several be combined?

Yes, and it is the usual thing: a hotel can have one program for the amounts, another for the guest chat and another for faults, from three different firms. What needs watching is the seam. When the guest assistant picks up a complaint about the room, somebody has to pass it to whoever fixes it. With Bedel that step is taken by a person: the front desk writes the fault to the hotel's number. Between a guest chat and Bedel there is no automatic connection, and it is worth bearing in mind when you draw up who does what.