A party of six walks in at seven-fifteen on a Saturday with no reservation, a host stand that is already three-deep, and a line forming on the sidewalk. The phone rings with another party asking what the wait looks like right now, and a catering inquiry for a forty-person rehearsal dinner in three weeks is sitting on the second line. Voicemail catches two of the three and the host has to triage the rest between seating tables. An AI receptionist for restaurants is built for the floor at seven-fifteen on a Saturday: it answers on the first ring, qualifies the booking in front-of-house vocabulary, and writes the reservation into OpenTable before the caller puts the phone down.
What missed calls cost a restaurant at the dinner rush
Most operators estimate that sixty to eighty percent of calls that hit voicemail during the dinner rush become tomorrow’s booking for somebody else — usually the OpenTable or Resy listing the caller pivots to on the walk back to the parking lot. A two-top Friday at seven PM clears two to four hundred dollars of contribution margin depending on the concept; a forty-person rehearsal dinner three weeks out lands four figures. Three lost calls on a busy weekend quietly erases a night of profit. Covering the host stand line for a flat per-month Starter plan recovers that margin from the first booking that would have pivoted to the chain down the road.
Front-of-house vocabulary on the first turn
A pre-trained restaurant model recognizes the request the way a host would — reservation, party size, waitlist add, takeout order, catering inquiry, high chair, allergies, occasion, dietary restrictions. It walks the caller through the structured intake a real host would run: name, party size, requested day and time, any flexibility, occasion (anniversary, business, birthday), notes on allergies or seating preference, and a callback number if the line goes long. Pulling a reservation out of voicemail and reading it on a clipboard has a half-life problem during the rush; getting it on the line, written into OpenTable before the caller hangs up, does not.
That last bit is what generalist answering services miss. They route the call or capture a message; the booking lives in a host's inbox until someone re-keys it the next morning. Hatchdesk writes the reservation into your real reservation system — OpenTable, Resy, SevenRooms, or Tock — with the right party size, time, occasion, and notes attached. The host reads a real booking on the floor at seven-thirty, not a forwarded transcript they have to chase down over a cortado.
TCPA-aware AI disclosure baked in
State voice-AI rules are tightening, and the failures are loud. Hatchdesk opens every call with a clear, TCPA-aware AI disclosure — never pretending to be a host — and logs the consent timestamp to your audit trail on every inbound call. The line is honest about what it is, and the record of who was told what is defensible against a regulator six months from now. Diners who pick up the phone again next week see the same disclosure; the record is consistent.
A flat per-month price is a feature, not a gimmick
Per-minute billing punishes you for being busy. A holiday weekend, a festival in town, a viral moment on TikTok — every spike is a bigger line item, which means the nights you most needed coverage are the most expensive nights. Hatchdesk is flat per month with per-minute transparency: one line item regardless of call volume. Predictable during the boom. Defensible against a slow Tuesday. And auditable against the bookings that came out of the line and the catering quotes that closed on the same shift.
FAQ: AI receptionists for restaurants
- Does the receptionist write a real reservation into OpenTable, Resy, SevenRooms, or Tock with party size, time, occasion, and notes attached — or does the host re-key a forwarded transcript on the clipboard?
- Is the model pre-trained on restaurant-floor vocabulary — reservation, party size, waitlist add, takeout order, catering inquiry, high chair, allergies, occasion — or is it a generalist with a few restaurant prompts bolted on?
- How is the TCPA-aware AI disclosure logged on every call, and can you pull the disclosure record for a specific call on request?
- What happens when a diner mid-call asks for a host or a manager, and where does the call route — the floor, the manager’s mobile, or the host stand?
- What does a Friday or Saturday busy weekend actually cost — flat per-month with per-minute transparency, or per-minute billing with add-on fees that punish your busiest weekends?