Restaurant owners track covers, spend per head and table turn. The number almost nobody tracks is calls that rang out during service — which is exactly when the volume arrives.
This is the arithmetic for that gap, plus what to fix regardless of what you buy.
Why the leak is concentrated in 4 hours a week
A restaurant's phone volume is not evenly distributed. It clusters:
- 17:00–19:00 — people deciding where to eat tonight
- Friday and Saturday evening — the same decision, with a booking attached
- Sunday afternoon — planning a group, a birthday, a work dinner
- Holiday run-ups — large bookings, questions about set menus and deposits
Those windows are also peak service. The person who would answer is carrying plates, seating a table, or taking an order. So the busiest phone hours coincide exactly with the hours nobody can pick up.
The arithmetic
Pick numbers and replace them with yours:
``
calls ringing out per week (Fri–Sun peak): 18
share that are booking enquiries (not suppliers/ staff): 70% -> 13
booking rate if answered: 65% -> 8 booked covers
average party size: 3.2
average spend per head: 120
= weekly revenue leak: 8 x 3.2 x 120 = 3,072
= monthly, at 4.3 weeks: ~13,200
``
Two nuances make it more significant than it looks:
Large bookings hide in the average. One 14-person birthday booking can be 1,700 in a single call. Those calls are longer, arrive outside service more often, and are the most likely to be missed because they happen at 14:00 when the phone is being ignored for prep.
A missed booking is rarely recovered. The caller is deciding tonight. They will book somewhere that answers. You do not get a second chance on Saturday at 19:30.
The four points where restaurants actually lose the call
1. The phone is in the wrong room. If the only handset is by the till, then during service it is a choice between the guest in front of you and the caller. The guest wins, correctly.
2. No answering system for the pre-service window. 15:00–17:00 is when most booking calls arrive and when staffing is thinnest.
3. Group bookings need information the phone cannot give at speed. Minimum spend, set menus, deposits, dietary handling, whether you can seat 14. Staff put these calls on hold, then forget them.
4. The voicemail box is full or not checked. This is the most common and the most embarrassing. A caller at 20:00 hears "the mailbox is full" and calls the next restaurant.
What to fix before buying anything
Measure first, for one week. Count:
- calls during service that rang out
- calls answered between 15:00 and 17:00
- group enquiries (6+) and how many converted
- calls between 22:00 and 09:00
If the out-of-hours number is zero and the peak-hour number is small, you do not have a phone problem — you have a staffing or layout problem, and software will not fix it.
If the numbers are large, the cheapest fixes come first:
- A second handset and a cordless in the service area. Free to 200, and it removes the worst of the collision.
- A written answer sheet for group bookings. Minimum spend, set menu price, deposit rule, capacity, dietary policy. Anyone can answer from it.
- A call-back discipline with an owner. Every unanswered call gets a return call within 20 minutes during the window. Track it or it will not happen.
Only after those are in place does an answering system make sense — and at that point you know exactly what it has to handle.
What an AI voice agent has to do to be worth it in a restaurant
The bar is concrete:
- Take a booking into the real table system, not a message. A message is a second phone call.
- Answer the four most common questions from your own material: hours, location and parking, dietary handling, and group policy.
- Recognise a group enquiry and route it to whoever handles deposits, with the party size, date, and contact already collected.
- Hand off anything about a complaint or an allergy to a person. An allergen answer invented by a system is a liability, not a convenience.
- Stop taking bookings when the kitchen is closed or the pass is full — which means knowing your actual capacity rules.
The honest constraint: a restaurant's knowledge changes constantly (specials, closures, seasonal menus, staff shortages). If nobody updates the knowledge base weekly, the agent will confidently book a table on a night you are closed.
Where this is the wrong purchase
- Walk-in driven venues where bookings are a small share of covers.
- Small rooms with fewer than ten calls a week — the subscription will cost more than the leak.
- Anywhere the owner wants every call personally. That is a relationship business, and automation erodes the thing that works.
- No reliable table management system. An agent that cannot write into a real calendar is just a better-dressed voicemail.
The two numbers to watch after
Bookings created outside service hours. This is new revenue, not displaced work — those calls were never going to be answered.
Share of calls resolved without a human. If it is low, the knowledge base is thin, not the voice quality.
Disclosure: I work on RingBot, an Agent-as-a-Service platform whose voice agents are built with an AI Builder, bound to your own knowledge base, and escalate to a person when a question needs judgement. The arithmetic above applies to any answering option — see the cost comparison of voicemail, answering services and AI agents and the same calculation for dental practices.
Top comments (0)