AI Booking Assistant for Restaurants (n8n + GPT)

I built a live booking assistant for a restaurant demo: a real floor plan, a live menu, and a chat AI that reserves tables, tracks party size throughout the conversation, and creates clear staff notes for details like allergies or special occasions.

Conversation state is preserved across multiple messages, and availability checks connect directly to the same database used for restaurant operations. Owner-side panels make it possible to update the menu and table setup live, with changes reflected by the bot immediately.

The bot runs on n8n with an OpenAI agent node at its core, and it’s designed to connect directly with the systems a business already uses — a real ERP or CRM, or a database with its own custom admin panel when that setup fits the business better. What you see in the demo is a fully interactive prototype: talk to the bot and watch reservations, notes, menu data, and other updates appear in the interface in real time.

How it works

The core of the system is an internal ticket-based CRM. The bot identifies the intent behind each request and tracks it as a separate ticket. That lets it handle several things at once — finalizing a reservation, answering a menu question, creating a note for staff, or continuing another open request — while keeping the context of each one clear.

Multi-intent tracking via tickets

Each request is identified and tracked as its own record, so the bot can manage several active conversation threads at the same time while keeping the correct context for each one.

Answers grounded in live restaurant data

Menu, table availability, opening hours, and policy questions are all answered through tool calls connected to live data, so the bot always reflects what’s actually in the system right now.

The bot remembers where you left off

Booking progress, guest details, and staff notes are stored as database records and refreshed on every interaction. A guest can return later and continue exactly where they left off.

Menu and table changes go live immediately

The menu, table layout, availability, and other restaurant data are managed through an admin panel, and the restaurant team can maintain all of it without touching the workflow itself.

Conversation as the interface

Booking happens through natural dialogue. The bot collects information step by step — party size, table availability, guest name — following the same flow a restaurant host would use.

Tested against real multi-turn conversations

The system was tested through complete conversations covering reservations, menu questions, changing requests, multiple simultaneous intents, and database updates, each checked against the actual records created and updated.

Stack

n8n with an OpenAI agent node, Supabase for persistent state and the admin panel, and REST APIs connecting the bot to live restaurant data.

Try it

A live version of the assistant is running at /chatbots/restaurant.