Use Case
AI Restaurant Reservation Host
Taking a reservation looks simple until the book is nearly full, the party is nine people, and one of them cannot sit near the kitchen. An AI persona built as a reservation host holds all of that in mind and never puts the phone down on a guest.
By EchoMP Editorial · Published 28 August 2026 · Updated 28 August 2026

The Scenario
A restaurant with a single dining room and a small private area puts the persona on its website beside the booking button and shares the same link in its social profiles and confirmation emails.
A guest opens it on a Tuesday afternoon wanting a table for six on Saturday at eight. The persona checks what is genuinely available, offers the nearest alternatives, notes the birthday and the one gluten-free diner, takes the contact details, and confirms the booking with the house policies said out loud.
How the Persona Is Set Up
Essentials
The role is a reservation host for one dining room: gracious, efficient, and honest about availability. The system prompt makes three rules non-negotiable — never confirm a table the system has not actually held, never bend the large-party or deposit rules, and always repeat the date, time, party size, and name back before finishing.
Knowledge
- Service times, last seating, and days the restaurant is closed
- Table inventory: sizes, seating areas, terrace, bar, and the private room
- Turn times per party size and how long a table is held for a late guest
- Large-party rules, deposits, set menus, and cancellation terms
- Accessibility, high chairs, pram space, and dog policy
- Dietary handling, corkage, dress expectations, and celebration options
Policies written as exact question-and-answer entries stop the persona from improvising a rule the floor manager will have to undo. See how knowledge and Skills work.
Guided Sessions
A "Book a table" session enforces the sequence: date, time, party size, seating preference, dietary or accessibility needs, name and phone number, read-back, confirmation. A "Change or cancel" session verifies the booking before touching it. A "Private dining enquiry" session collects the brief and routes it to the events manager.
Skills
- Availability lookup against the live reservation system
- Booking creation, amendment, and cancellation
- Waitlist entry with a callback when a table opens
- Confirmation and reminder messages to the guest
- Deposit request for large parties where the restaurant requires one
- Handoff to the events manager for private hire and set menus
Say No Gracefully
The skill that matters most on a full night is the refusal. Rather than a flat "we are fully booked", the persona offers the two nearest times, the bar counter, or a place on the waitlist with an honest chance of it moving.
It never holds a table on a maybe, and it never quietly overbooks to please the person in front of it. A booking the floor cannot honour costs more than a booking never taken.
Handling the Awkward Moments
- A party grows from six to ten after booking — it re-checks availability rather than assuming
- A guest disputes a cancellation fee — it states the policy and offers a manager
- A serious allergy is mentioned — it records it on the booking and flags the kitchen
- Someone wants to buy out the room — it collects the brief and routes it to events
- A complaint about a previous visit — it hands off to a person straight away
What the Restaurant Gets
- Bookings taken at the exact moment the guest decides, including outside service hours
- Complete records with dietary notes and celebrations attached
- Fewer no-shows, because confirmations and reminders always go out
- A working waitlist instead of lost enquiries on a full Saturday
- Hosts on the floor greeting guests rather than pinned to the phone
The same persona can be embedded on the restaurant's site or shared as a link — see share, embed, meetings, and earnings.
Related Use Cases
Never miss a booking on your busiest night.
Load the floor plan, the policies, and the service times, and let every enquiry end in a confirmed table or an honest alternative.
