Reservations and Allergens: AI Chat in Restaurants

Friday evening, just before seven: the kitchen is in full swing, orders pile up at the counter, and the phone rings for the fifth time. On the other end, someone simply wants to know whether there is still a table for four at half past seven. At that very moment no one is free to pick up, and the request ends up on the answering machine or at the place next door. For restaurants and hotels this is daily life: reservation requests, questions about opening hours, the menu, allergens or free rooms often arrive exactly when there is the least time for them. The pressure is growing, because the shortage of skilled staff is still felt in the industry. An AI chat assistant on the website takes on precisely these recurring requests. It answers questions in normal language, holds reservations and stays reachable even when the team is with the guests rather than on the phone. This article shows which tasks an assistant handles in a restaurant and a hotel, where its limits lie and how service and phone can be noticeably relieved, without guests having to give up personal contact.

Enlarge graphic

Labels in the illustration

• XICBOT
• Table for 4 on Friday, 7:30 pm?
• Sure. Friday 7:30 pm is free —
• shall I request the table for you?
• Do you have gluten-free dishes?
• Yes, several main courses are
• marked as gluten-free.
• Reservation requested
• Friday 7:30 pm · 4 guests
• Type a message...
• Free times from the stored time windows
• Reservations, questions and allergens around the clock
• The assistant answers
• Opening hours and directions
• Menu and allergens
• Reservation and table request
• Rooms, rates and availability
• Why now
• 71%
• of weekly guests
• have already booked
• a table online
• (YouGov)
• Staff shortage
• still felt in many businesses
• Answer reservations, opening hours, menu and allergens automatically

AI Assistant for Restaurants and Hotels

Why the Phone Becomes a Bottleneck in Hospitality

The phone is a serial channel: it handles only one conversation at a time. During the evening peak, when kitchen and service are already stretched, this turns into a queue, a busy signal or a callback that often does not happen. Every one of these moments is a point where a reservation is lost. For the business this means not only empty tables but also tied-up staff. The bottleneck is real: many restaurants still feel the shortage of skilled workers. At the same time the number of VAT-registered hospitality businesses fell by 7.3 percent (DEHOGA Bundesverband) between 2019 and 2024, while the number of employees subject to social insurance, at just over 1.12 million (DEHOGA Bundesverband) in September 2025, was slightly above the 2019 level. Anyone with fewer hands can least afford one of those hands being permanently stuck on the phone.

On top of that comes the time limit. A large share of requests arises outside staffed hours, late in the evening during weekend planning or on a Sunday morning when no one is at the phone (project experience). Guest behaviour has shifted too. People are used to sorting things out themselves and immediately: according to a representative survey, 70 percent (YouGov) of guests use digital platforms to find restaurants, and 50 percent (YouGov) have already booked a table online, rising to 71 percent (YouGov) among weekly diners. Since a large share of page views in Germany come from mobile devices – 44.92 percent (Statcounter) in February 2026 – the potential guest is often already on the page and simply waiting for a quick answer. An assistant in the chat closes this gap by adding a parallel, round-the-clock channel to the serial, staffing-bound phone intake.

What an AI Assistant Handles in the Restaurant

A good assistant mirrors the everyday work at reception and phone rather than being a technical toy. It answers the same recurring questions reliably from your own content and hands over the rest to people. The benefit lies less in the single conversation than in the sum: when routine requests run through the chat, the team gains time for the guests on site. This relief matches a broader pattern in customer service. In a McKinsey maturity model for AI-enabled customer service, self-service channels such as apps, voice response systems and websites handle 70 to 80 percent (McKinsey) of interactions at the second-highest level, and in a bank example described by McKinsey the number of service interactions fell by 40 to 50 percent (McKinsey) after its customer service was rebuilt. Applied to a restaurant or hotel this means: the questions that block the phone today can largely be cleared in the chat.

Take Reservations Around the Clock

The reservation is the core. An assistant that only answers questions leaves the most important step open; an assistant that also takes reservation requests leads the conversation to a result. The flow feels like a short, friendly dialogue and still leads to a clean hold. What matters is where the assistant gets availability from: it calculates free times from the time window, duration and buffer of the stored appointment type minus reservations already requested or confirmed and, when the house is full, suggests alternatives from them, such as an earlier sitting or another evening. It records the wish as a request and the restaurant confirms; a binding direct entry happens only through a set-up calendar integration (Custom package), and a table plan or reservation system can be connected as your own interface via tool control. How such a flow with confirmation and reminder can be implemented in detail is described on the page about the booking assistant.

• Greeting that suits the page and a question about date, time and party size
• Check the desired time against the stored time windows and reservations already requested, offer free slots and record the wish as a request
• Capture special requests, such as terrace, high chair or step-free access
• Hold the reservation as a request the restaurant confirms; enter it directly and bindingly only through a set-up calendar integration (Custom package)
• Remind in good time before the confirmed date, if an email address is on file
• For special cases such as large groups or events, hand over to the team

In this way a single assistant covers the path from the first question to the held reservation, late in the evening and at the weekend too. It replaces not only the enquiry form but also the manual follow-up: recording the request and, if an email address is on file, reminding shortly before the confirmed date. How the conversation around the reservation request feels in everyday use and which subtleties matter is explored in the article Appointment requests via chat assistant. Which reservation requests are especially common and where guests drop out of the conversation is made visible by the conversation analytics.

Allergens and Menu: Accuracy Matters Here

With the menu an assistant is helpful, with allergens it is delicate. Questions like “Is the risotto gluten-free?” or “Do the dumplings contain egg?” must not be answered with something made up. That is why a properly set up assistant binds its answers strictly to your stored details about dishes, ingredients and the fourteen allergens that require labelling. Where a detail is missing or unclear, it does not guess but says so openly and points to the kitchen or the service team. This binding to checked sources is not an accessory but the basis for making sure convenience does not turn into misinformation. How to prevent an assistant from inventing freely is explained in the article Preventing hallucinations with a knowledge base.

The assistant’s knowledge comes from a well-maintained knowledge base: your menu, ingredient lists, opening hours, house rules and typical questions. If the menu changes, the answer changes as soon as the source is updated. That keeps the assistant on the level of your real offering rather than following an old printout. Precisely because digital menus are on the rise – 38 percent (YouGov) of guests have already used them – it pays to maintain this content cleanly and make it available to the assistant.

Hotels: Room Questions and Enquiries Before Booking

In the hotel world the focus shifts from the table reservation to the room enquiry. Before a guest books, they usually check a few details: is a double room free on the desired weekend? What does the night with breakfast cost? Are a dog, parking or a late arrival possible? An assistant answers these questions from your details, states availability and rates and, given concrete interest, leads to the enquiry or the regular booking path. Because many guests travel from abroad, a multilingual assistant helps, answering the same question in German, English or further languages without anyone at reception having to speak them. How this pays off internationally is shown in the article Multilingual AI assistant.

Around the stay further opportunities arise. An assistant can point to add-on services – a breakfast upgrade, a voucher or a package – and place these offers in the conversation without becoming pushy. Where a hotel runs a small shop for vouchers or packages, the same principles apply as in retail: clear open questions early, do not overload the guest with offers and lead them to completion at the right moment. Which subtleties apply is described in the article Recovering cart abandonment with a shop chat. This way an open enquiry more often becomes a concrete booking.

A Conversation in the Restaurant, Step by Step

An example makes what this feels like in everyday use tangible. In the evening a guest opens the website and types into the chat: do you have a table for four on Friday at half past seven? The assistant understands this perfectly ordinary phrasing without anyone having to click through a form. It checks the stored time windows and the reservations already requested, sees that 7:30 pm is free and asks whether it should request the table. Midway, the guest wants to know whether there are gluten-free main courses. The assistant answers from the labelled menu and then leads back to the reservation. In the end there is no open question in an inbox but a held reservation request that both sides have in black and white and that the restaurant only needs to confirm. It is exactly this understanding of natural language that sets a website assistant apart from a rigid reservation widget.

When It Gets Personal: Handover to the Team

A good hospitality assistant replaces no hosts, it relieves them. It takes on the recurring, well-structured work and hands over to the team wherever things get personal, sensitive or unusual: with a complaint, a large celebration, a special request that does not fit the pattern or a delicate allergy question. The right answer is then not a forced booking attempt but a clean handover with full conversation context, by email or straight to a member of staff. How this handover succeeds smoothly, without the guest having to repeat their question, is described in the article AI assistant: handover to human agents; in everyday use this meshes closely with the support assistant.

Connect Securely: Table Plan, PMS and Privacy

So that the assistant holds reservations with a specific time as a request instead of merely collecting contact wishes, it needs controlled access to clearly outlined functions. Technically this happens through defined actions, often called function calling: the assistant may call clearly outlined functions, such as naming free times from the stored appointment types or creating a reservation request. It gets no blanket full access but only the rights needed for the respective task. How this interplay of reading and controlled action is built is described on the page about tool control. Depending on the business, a voucher shop is connected or, for binding direct entry, a calendar through a set-up calendar integration (Custom package); a table plan, reservation or property management system is added as your own interface.

That this investment pays off also has an economic side. Costs in hospitality have risen sharply: labour costs alone were 39.6 percent (DEHOGA Bundesverband) above the end-of-2019 level in the fourth quarter of 2025. In a market with 206,105 VAT-liable businesses (DEHOGA Bundesverband) and a net turnover of 118 billion euro (DEHOGA Bundesverband) in 2024, every request that is not lost counts. Because personal data such as name, contact and reservation request are processed, privacy and hosting belong in from the start: hosting in Germany, a data processing agreement and a clear deletion policy. Details are set out on the page about privacy and hosting. That shifting routine to digital channels nonetheless brings measurable benefit is shown by a look beyond the industry: 42 percent (McKinsey) of leading service organisations managed to reverse rising request volumes through smart self-service.

• Store opening hours, menu and allergens in a well-maintained knowledge base
• Store time windows, set up a calendar integration with permissions for binding direct entry (Custom package), and the table plan or reservation system as your own interface where needed
• Plan for confirmation and ask for the email address for the reminder, to reduce empty tables
• Define the handover to the team for complaints, large groups and delicate cases
• Secure privacy and hosting in Germany by contract
• Keep refining through the conversation analytics which questions come up most often

To Your Own Hospitality Assistant in a Few Steps

The path to your own assistant begins with a short look at your most common requests, your menu and your reservation flow. On this basis the assistant is trained individually, connected to your system and embedded in the website with a short snippet. After that it takes reservation requests and answers questions while your team looks after the guests on site. A fixed rate cannot seriously be promised, because how many enquiries turn into reservations depends on offering, location and website. What is dependable, by contrast, is the mechanism: reachable around the clock, an answer in seconds, confirmation by the restaurant and, if an email address is on file, confirmation email and reminder included. An overview is given by the services and the industry page on the AI assistant for restaurants and hotels; if you would like to run this through your own case, you can request a free demo for your restaurant or hotel. Where a structured enquiry rather than a firm reservation makes sense, the lead assistant steps in as well.

Frequent questions