WhatsApp Chatbot for Restaurants: Orders and Bookings

A WhatsApp chatbot for a restaurant is a program that answers your customers' messages automatically: it sends the menu, takes a delivery or takeaway order, books a table, confirms it, and passes everything to the kitchen or to your booking system. It works well when you get a steady flow of repetitive messages — opening hours, delivery area, "do you have a table for four at 8". It works badly when every conversation is different, or when a person at the counter already answers within a minute. Below is what it actually involves and how to decide.

What the bot does in practice

A guest writes to your WhatsApp number. The bot replies instantly with a short menu of options: order delivery, book a table, see the menu, talk to a person.

For an order, it walks through the dish list, quantities, address, delivery time and payment method. At the end it sends a summary, the total, and a confirmation. The order lands where your team already works — a printer in the kitchen, a Telegram group, your POS, or an order screen.

For a reservation, it asks date, time, number of guests and name, checks availability against your table plan or booking tool, and confirms or offers the nearest free slot. It can send a reminder a few hours before and let the guest cancel with one tap, which cuts no-shows without anyone making phone calls.

Everything the bot cannot handle goes to a human. That handover is the most important part of the build, not an afterthought.

When it makes sense

A chatbot pays off in a restaurant when at least one of these is true.

• You get more than a few dozen messages a day and someone is reading them between shifts.
• A large share of messages are the same five questions.
• You lose bookings in the evening because nobody picks up the phone during service.
• You run delivery or takeaway directly, not only through aggregators, and orders arrive as free-form text you have to retype.
• You have several locations and guests write to the wrong one.

It also makes sense when you want your own customer list. Orders taken through an aggregator belong to the aggregator. Orders taken on your own WhatsApp give you a phone number, an order history and a reason to message that guest again, within the rules of consent.

When it does not

A small place with one room, a full house every night and a host who answers the phone in ten seconds does not need a bot. It needs a clear menu page, a click-to-WhatsApp button and correct opening hours on Google Maps. That is a much cheaper fix.

A restaurant with no delivery and a table plan kept in a paper notebook usually should not start with AI. Start by putting the bookings in one shared system that the whole team can see. Automate after the process exists.

If your entire volume comes through delivery platforms and you have no intention of taking direct orders, a chatbot adds a channel nobody uses. Fine-tune the platform listings instead.

And if your menu changes daily and by hand, be honest about who will update the bot. A bot serving last week's menu is worse than no bot.

Scripted flows versus an AI chatbot

Most restaurant work is better served by a scripted flow: buttons, fixed steps, predictable results. Scripts do not invent dishes, do not promise a table that does not exist, and are cheap to test.

An AI layer on top helps with the messy edges: free-text questions about allergens, ingredients, parking, dress code, or a guest who writes "table for 6 Friday night, one high chair" in a single sentence. The safe pattern is AI for understanding and answering questions, scripts for anything that creates a booking, a bill or an order. Give the AI only your real documents — menu, allergen list, house rules — so it answers from facts instead of guessing.

What to connect it to

A chatbot in isolation is a toy. The value comes from the connections.

• Menu source. One place where prices and availability live, so the bot and your website show the same thing.
• Booking system or table plan. Otherwise the bot double-books.
• POS or kitchen display, so orders do not get retyped.
• Payments, if you take prepayment or deposits for large groups.
• CRM or a simple customer database, so repeat guests are recognised and you can see who ordered what.
• Your team's channel — Telegram, a shared inbox — for handover and for the cases the bot refuses.

Use the official WhatsApp Business Platform rather than an unofficial workaround. Unofficial connections get numbers blocked, and losing your restaurant's main number in the middle of a Friday is an expensive way to save money.

What to prepare before you start

Write down the twenty questions your staff answer most often, with the real answers. Export your menu with prices, sizes and allergens. Decide your delivery zone and minimum order. Decide who is on duty to pick up handed-over chats and within how many minutes. Decide what the bot must never do — cancel a paid order, give a discount, confirm a group of twenty without a human.

That document is most of the project. Contractors who skip it are guessing.

How to choose who builds it

Ask to see a working bot, not a slide. Ask how the handover to a human works and how the bot behaves when the booking system is down. Ask who can edit the menu and the answers after launch — if only the developer can, you will pay for every price change. Confirm the WhatsApp number stays yours and that you get access to the accounts.

Start with one job. Reservations only, or takeaway only. Run it for a month, count how many conversations it finished without a human, then extend.

What it costs

Public starting prices: automation and AI agents from 250 000 ֏, integrations with an existing system from 149 000 ֏, support plans after launch from 9 900 ֏ per month. These are starting points; the exact figure follows a short brief, because a button-based booking bot and a multi-location AI assistant tied to a POS are different projects. Budget separately for WhatsApp conversation fees charged by the platform itself.

Frequently asked questions

Will guests notice they are talking to a bot? Yes, and that is fine if it is fast and honest. Say it is an assistant, offer a human in the first message, and never let a guest loop.

Can it replace a reservation phone line? It can absorb most routine bookings and reminders. Keep the phone for large groups, events and complaints.

How long does a chatbot for a restaurant take to build? It depends on the integrations, not on the bot. A scripted reservation flow is quick; connecting a POS and payments is what adds time. Ask for a timeline in writing after the brief.