Hospitality
AI in the hospitality industry: what it does today in a restaurant or hotel
Questions, bookings, menu copy and rotas prepared by AI; the decisions and the welcome stay with your team.
The hours between services
AI in the hospitality industry mostly does work between services: answering questions and booking requests, writing menu copy, replying to reviews and forecasting next week's covers for the rota. The monthly bill is usually small; the work is in setting it up well.
A restaurant or hotel gets the same questions every day: is there a table on Saturday, is there parking, can my daughter eat gluten free, can the dog come. In English, French, German and Dutch. Whoever answers is not on the floor or in the kitchen at that moment.
This page sets out six uses with what they cost a month and what they save, the line guests draw, what allergen rules mean for AI, and how to start. Up front: we have no restaurant as a client; the proof comes from booking and messaging systems we built.
Six uses and their monthly cost
Model costs in hospitality are almost always small, because the texts are short. What counts is the setup: which questions the assistant may answer, from which information, and when it hands over to a person.
| Use | What AI does | Cost a month, roughly | What it saves |
|---|---|---|---|
| Questions and bookings | Answers email, chat and messages in four languages, offers a table or room | Under 1 dollar of model cost for 1,000 questions on a light model, plus the link to your booking system | An hour or more a day of messages |
| Phone after hours | A voice assistant that takes the request or offers a booking | About 0.05 dollars a minute for the voice model, plus the language model | Missed calls that never became a booking |
| Menu copy and allergens | Descriptions of new dishes in four languages, with allergens from the recipe card | A business plan from about 21 euro per seat a month | An evening per seasonal menu |
| Reviews | A reply to every review in the guest's language, ready to check | A few cents | Replies that would otherwise not get written |
| Covers forecast | Covers per hour for next week from your till, bookings and the weather | A one-off build, little after | A rota that fits better |
| Purchasing | An order list from sales and bookings | A one-off build, little after | Less waste and fewer rush orders |
Model prices are the vendors' own, in dollars per million tokens; a light model such as Claude Haiku 5.5 costs 0.10 dollars per million input tokens and 0.50 output. Check your own volume with the AI cost calculator; what a chatbot costs in full is on AI chatbot cost.
A day in a restaurant, with what AI prepares
Purchasing
AI preparesAn order list from yesterday's sales and today's bookings
You decideThe chef adjusts and orders
Reviews
AI preparesA reply to every new review, in the guest's language
You decideRead, edit, post
Menu
AI preparesDescriptions of new dishes, in four languages, with the allergens from the recipe card
You decideThe chef checks every allergen
Social
AI preparesA post about the special of the week
You decidePick a photo and approve
Rota
AI preparesExpected covers per hour for next week
You decideThe manager sets the shifts
Bookings
AI preparesAnswer questions by email and chat, offer a table
You decideAnything outside the rules
Daily report
AI preparesRevenue, covers and what stood out, in five lines
You decideRead it with the morning coffee
Example of a restaurant with thirty covers per service. The business stays responsible for allergen information.
Examples per kind of business
The same technology carries a different weight in each business.
| Business | What earns the most | Where it goes wrong |
|---|---|---|
| Restaurant | Bookings and questions outside service hours, allergens on the menu, reviews | An allergen not on the recipe card, a group outside the rules |
| Hotel | Pre-arrival questions in several languages, upgrades and extras, reviews on the platforms | Confirming a room that is already gone, when the system is not connected |
| Bed and breakfast | Questions about availability and the area, preparing a booking request | A window between calendars in which the same night sells twice |
| Catering | Quotes for events from a set menu, with numbers and dietary needs | A price that does not hold for a large number |
In every case the assistant is as good as the information you give it: an up-to-date menu, your rules for groups and your opening hours matter more than the model. How assistants answer and hand over is on AI customer service; a chatbot on your own site is on AI chatbot development. For rooms without a restaurant, see bed and breakfast booking software.
The line guests draw
Guests are fine with an assistant that tells them at eleven at night that there is a table on Saturday. They are not fine when nobody answers a complaint, or when an answer about their allergy is wrong. So each use gets rules for what the assistant does alone and when it hands over.
- Answers aloneopening hours, parking, dogs, children's menu, availability, directions.
- Proposes, a person confirmsa group booking, a menu for a party, a room with a special request.
- Always to a personcomplaints, allergies not on the recipe card, payments and refunds, anything personal.
An assistant also says it is one: the EU AI Act asks that of a chatbot talking to people, and it keeps a guest from feeling misled. How to set that down for your team is in the EU AI Act for business and do I need an AI policy.
Our booking and guest flows, read for a restaurant
For Villa Lidia, a villa in Umbria, we built a site in two languages with an availability calendar, a booking form and a small admin panel. For No.4 Lange Ast, a calendar from the Airbnb and Booking.com iCal with a request form in four languages.
For DA Logistics we built a mail layer with fourteen message types that go out by themselves from what happens in planning, and an app in which drivers send reports and questions in three languages. It is the same stream as guests: questions coming in, answers going out on time, a person keeping the overview.
An assistant for a restaurant puts those building blocks together: your booking system, your menu with allergens and a language model that answers within your rules.
A worked example for a restaurant
The setup, the link with your booking system and teaching it your answers are the real cost; we quote a fixed price per project after a first call. How a business brings AI in step by step is on AI for business and implement AI in business. A site guests find and book on directly is on restaurant website, and the rota itself on staff scheduling software.
AI for business
How VBS helps you with this
We build AI into your own work: it drafts emails, reads documents and answers questions, and you stay the one who decides.
- First a task that costs time, then a working prototype
- On your own documents and your own data
- An AI policy your team understands
How we approach it is on our page about AI for business.
AI that takes load off without pushing guests away?
Tell us how many messages you get a day, in which languages, and which booking system you use. After a call you get a proposal for a first use and a fixed price.
Rather get in touch right away?