Metaviz AI
AI Automation SaaS

A voice agent that answers the phone and books the table

Restaurant Reservation Voice Agent

A voice agent that answers a restaurant’s incoming calls, books, moves and cancels reservations, and writes every booking into Google Calendar and a customer database without anyone typing it in.

Overview

Restaurants lose a good part of the day to the phone. Someone stops what they are doing, takes a name and a time, writes it in a book, and hopes the next person can read it. We replaced that with an agent that answers the call itself.

The agent handles the whole conversation. How many people are coming, what date and time, whether that slot is actually free, and anything the table needs. If the caller has not been in before, it takes a name, an email and a phone number. It then books the table, creates the calendar event, and stores the customer record.

Retell AI runs the voice layer. n8n sits behind it and moves the booking into Google Calendar and Supabase the moment the call ends, so there is no list of bookings waiting to be typed up later.

Restaurant Reservation Voice Agent - project overview
- Relevant keywords -
  • Voice AI Agent
  • Restaurant Reservations
  • Retell AI
  • n8n Automation
  • Google Calendar Sync
  • Supabase
  • Inbound Call Handling
  • Booking Automation
Challenge & Solution

Solving real problems with smart engineering

Staff answering the phone instead of working the floor

Problem

Every reservation meant stopping a task, taking details by hand, and writing them into a book.

Solution

The agent answers every inbound call and takes the booking from start to finish, so nobody has to leave the floor to do it.

Two tables promised the same slot

Problem

A paper book and a busy service is how the same time gets given to two different people.

Solution

Before it confirms anything the agent checks live availability on the calendar, so a slot can only be sold once.

Customer details spread across three places

Problem

Names and numbers ended up in the booking book, in somebody’s phone, and in nobody’s memory.

Solution

Each caller becomes one record in Supabase, and the same record is recognised and updated the next time they ring.

The calendar kept by hand

Problem

Bookings had to be copied into the calendar after the call, which is where a fifth to a third of admin time went.

Solution

n8n creates the calendar event the moment the booking is confirmed, with the exact time on it.

Special requests going missing

Problem

Allergies, birthdays and seating preferences were written in the margin of a page, when they were written at all.

Solution

The agent asks for them during the call and saves them with the booking, so the kitchen sees them.

How it works

What happens during a call

The agent follows the same order every time, which is what keeps the data clean at the end of it.

Step 1

The call comes in

The agent answers and greets the caller.

Step 2

Party size

It asks how many people the table is for.

Step 3

Date and time

The caller gives the day and the time they want.

Step 4

Availability check

The agent checks the calendar for that slot before going any further.

Step 5

Special requests

Dietary needs, celebrations and seating preferences are captured here.

Step 6

Caller details

A first-time caller is asked for name, email and phone.

Step 7

Customer record

An n8n webhook creates the customer record.

Step 8

Booking

The reservation is booked against the exact timestamp.

Step 9

Calendar

n8n creates the matching Google Calendar event.

Step 10

Database

n8n writes the booking metadata to Supabase.

Features

What it does

On the call

  • Answers inbound calls and greets the caller
  • Takes party size, date and time
  • Checks live availability before confirming anything
  • Books, reschedules and cancels reservations
  • Captures dietary needs, celebrations and seating requests
  • Collects name, email and phone from first-time callers

After the call

  • Creates the Google Calendar event with the exact timestamp
  • Writes the customer record and booking metadata to Supabase
  • Sends the caller a confirmation message
  • Runs a post-call analysis pass over the transcript
Gallery

Inside Restaurant Reservation Voice Agent

Tech Stack

Built with

Voice

  • Retell AI
  • 11Labs (Cimo voice)

Conversation

  • GPT-5 for dialogue logic
  • GPT-4o-mini for post-call analysis

Automation

  • n8n webhooks
  • Availability lookup workflow

Calendar

  • Google Calendar API

Data

  • Supabase
Results & Value

What actually changed

The phone stopped being a job. Every call is answered, every booking lands in the calendar and the database on its own, and the front of house gets its attention back.

  • Reservations taken without a member of staff on the phone
  • One customer record per caller instead of details in several places
  • Special requests recorded with the booking rather than in the margin
  • Confirmation sent to the caller without anyone sending it
01Headline outcome
100%Calls handledEvery inbound reservation call is taken by the agent.
02
99.2%
Complete customer recordsName, email, phone and any special request captured on the call.
03
90%
Less admin timeBookings no longer have to be written up and copied into a calendar.
04
0
Double bookingsAvailability is checked against the live calendar before a slot is confirmed.
05
100%
Bookings syncedCalendar event and database record created on every confirmed booking.

Have a similar project in mind?

Let's talk about how AI can transform your service business.

Talk to us about a system like this