The phone rings at 19:20
Nobody can leave the floor to answer it. The caller books somewhere else and does not ring back.
See how the sector uses AIDieses Beispiel wird gerade überarbeitet. Der Text unten ist noch auf Englisch.
Example · Restaurants and cafésReservation and prep agent
The phone rings during service, and nobody can reach it
An agent that answers the booking line, takes the details you need for the table plan, sends the confirmation, and puts every allergen question in front of the kitchen.
Illustrative scenario — not a real client result
Booking line · 19:20
- Answered
4 bookings taken
19:21 · Party size, time, highchair, notes
- Answered
Two tables confirmed
19:23 · SMS confirmation, free cancellation until 16:00
- Person
“Is the curry dairy-free?”
19:24 · Allergen · kitchen answers, not the agent
No allergen answer is ever generated
Reservation and prep agent · Restaurants and cafés
- Sourced facts
- The kitchen decides
- Bookings only
01 — A typical day
Three small failures that add up.
What restaurants and cafés owners tell us goes wrong in an ordinary week.
An allergen question answered by guesswork
“Is there dairy in this?” A confident answer from anyone who is not the kitchen is a risk, not a courtesy.
See the inquest findingTables that never turn up
A no-show at 20:00 on a Friday is a lost cover you cannot resell at 20:20.
See the workflow01 — A typical day · the evidence
What the numbers actually say.
Each figure opens its source, year, sample and limits. Click a badge to check it.
01
Source
Deloitte, AI in restaurants report
2025
Method and sample are recorded in the source registry.
Read the source for its scope and limits.
Open the source63%
of surveyed restaurant leaders use AI daily for customer experience
Deloitte, 375 restaurant leaders, fieldwork late 2024. 93% of respondents worked at organisations with 1,000+ staff — not an independent-restaurant figure.
Deloitte · 202502
Source
Deloitte, AI in restaurants report
2025
Method and sample are recorded in the source registry.
Read the source for its scope and limits.
Open the source55%
use AI daily in inventory management
The same Deloitte survey. A large-operator statistic, labelled as such on this page.
Deloitte · 202503
Source
Deloitte press release on AI in restaurants
2025
Method and sample are recorded in the source registry.
Read the source for its scope and limits.
Open the source82%
expect their organisation's AI investment to rise next year
A stated intention, not realised spending, and again weighted to large chains.
Deloitte press release on AI in restaurants · 202502 — What the helper does
From your records to a decision you own.
The agent answers the booking line and the web form, takes party size, time, occasion, accessibility and contact details, and offers only the slots your table plan actually has.
It confirms by SMS, states the cancellation rule, and adds anything the kitchen needs to know to the prep list — except allergens, which are routed to a person.
A request for a large party, a private hire, a refund or an allergen answer goes to you. The agent never improvises on food safety.
6 steps · 2 decisions you own
Reads from
- Booking line
- Web and app form
- SMS
- Table plan
Text version of the flow
- 1. Take the booking — Party, time, occasion.
- 2. Check it fits — Capacity and turn time. Human checkpoint: You set the table rules.
- 3. Confirm and set the rule — Hold, cancel, arrive.
- 4. Route the allergen question — Straight to the kitchen. Human checkpoint: The kitchen answers allergen questions.
- 5. Update the prep list — Covers, notes, occasions.
- 6. Remind and release — Day-of confirmation.
02 — What the helper does · a simulated example
One simulated example, step by step.
Use the buttons or the left and right arrow keys. Nothing here is a real customer.
Simulated example
The Copper Spoon, a 48-cover bistro
Booking line and SMS · 19:20
A single-site bistro at Friday service. The agent answers the booking line while the floor is full. Every message is simulated; no real guest data is used.
The Copper Spoon's booking assistant here. I can take a reservation or pass a message to the team. Reply HUMAN for a person.
03 — What you decide
The helper prepares. You decide.
The agent takes bookings. The kitchen owns the food.
Allergens stop the thread
The question goes to a named person with the dish and the booking attached; the guest is told who will answer.
Your table rules
Sittings, party sizes, hold times and deposits are yours. The agent offers nothing outside them.
A person takes over
Large parties, private hire, complaints and refunds go to you, with the thread.
You can pause it
Pause the line for a service, a day or a closure; callers hear your own message.
Without your approval · No ingredient, allergen or dietary claim is ever generated, confirmed or implied by the agent — not even to be helpful.
04 — What changes when a safeguard is missing
Documented cases, not scare stories.
Three documented incidents with their sources. The first is why allergen questions never reach the agent; the others are why bookings need an audit trail and a fallback.
A verbal note the kitchen never saw
Owen Carey told staff he had a dairy allergy. He was not told his grilled chicken contained buttermilk, and the coroner found the menu did not mention the marinade. He died after a severe reaction; the inquest reported in 2019.
What the helper would do differently
The agent's only allergen output is a routed question and the name of who will answer. The dish, the booking and the query travel together, so a verbal note cannot be lost between the door and the pass.
1 death
anaphylaxis after an unlisted buttermilk marinade
Three hundred fake bookings from one account
An OpenTable employee made roughly 300 fake reservations on a rival service. Reporting said 45 restaurants were affected and more than 1,200 no-show guests were recorded, including around Valentine's Day.
What the helper would do differently
Bookings are written with an account, a timestamp and a source, and unusual volumes on one channel are flagged rather than absorbed into the plan.
1,200
no-show guests reported across 45 restaurants
A configuration change stopped the tills
A third-party provider's configuration change triggered a global technology outage that stopped affected restaurants taking orders. One UK franchisee said all 21 of her branches could not serve anyone for about 90 minutes.
What the helper would do differently
The booking line is not the till. If the agent is unavailable, the number falls back to the restaurant and a paper booking sheet, and the plan reconciles when systems return.
90 min
one franchisee's 21 branches unable to serve
04 — What changes · the evidence
The rest of the evidence.
Every figure resolves to a source with its sample and limits. No figure without one.
63%
of restaurant leaders use AI daily for customer experience
Deloitte, 375 leaders, late 2024. Weighted to organisations with 1,000+ staff.
Source
Deloitte, AI in restaurants report
2025
Method and sample are recorded in the source registry.
Read the source for its scope and limits.
Open the source375
restaurant leaders in the Deloitte survey
Fieldwork October–November 2024, 11 countries. Method beyond sample size is not published.
Source
Deloitte, AI in restaurants report
2025
Method and sample are recorded in the source registry.
Read the source for its scope and limits.
Open the source55%
use AI daily for inventory
Same survey. Included to show the pattern is operational, not only marketing.
Source
Deloitte, AI in restaurants report
2025
Method and sample are recorded in the source registry.
Read the source for its scope and limits.
Open the source
04 — What changes · your week
The same week, two ways.
Drag the handle, or use the left and right arrow keys when it is focused.
A Friday, beforeA Friday, after
19:00
The phone rings four times during the first turn; three callers hang up.
19:30
A waiter answers an allergy question from memory while carrying plates.
20:00
Two tables are no-shows; the plan was never confirmed.
22:30
Tomorrow's covers are added up from a paper pad.
19:20
Every call answered; bookings land on the plan as they are made.
19:24
Allergen questions are with the kitchen, each tied to a booking.
16:00
Day-of reminders went out earlier; cancelled tables were released.
22:30
The prep list is already current, with covers and notes.
Illustrative example · not a measured result
05 — Your version
Four questions about your own week.
Keep your answers. They are the brief for the first version we would build for you.
How many booking messages arrive in a week, and how many come during service?
What do you need to know for the table plan, and what is too much to ask?
How many no-shows or late cancellations does a normal week have?
Who answers an allergen question, and who confirms the answer?
Next
Put your own figures into the calculator below. It runs in your browser, and nothing is sent or stored.
Put your week into the calculatorBring the four answers to the free one-hour call. We will use them to sketch the first version, and you keep the written plan either way.
Talk about your business05 — Your version · your numbers
Put your own week into it.
An estimate, not a promise. The model and its assumptions are printed below.
Your inputs
Reservations, not walk-ins.
count
Your honest estimate for a busy Friday.
%
Phone, notes, confirmation, changes.
min
Whoever covers the phone.
per hour
From your own booking log.
%
Example defaults are typical for a small business. Replace them with yours. Nothing is stored, sent or shared.
Estimated result
IllustrativeFloor hours back each month
0–0
- Time released at your rate
- £0–£0
- Calls answered during service
- 0–0%
- Time to a confirmation
- Already immediate
How we calculated this
Routine work handled by the helper: 70% of the share you marked routine.
Time saved per item: 80% of the minutes you entered.
Monthly volume: weekly figures × 4.33; monthly figures used as entered.
Cost: hours saved × your staff cost per hour, in pounds as an illustration.
Reply time: routine items move within the hour while the helper runs. Complex work still waits for a person.
Questions
The things people ask next.
No. It states that the kitchen will answer, and routes the question with the dish and the booking. It never confirms what a dish contains.
Only within the party size you set. Anything larger goes to you, with the details already collected.
It can sit in front of the platform you use, reading availability through its API. Direct bookings stay direct.
The line falls back to the restaurant and a paper sheet. There is a written reconciliation step for when systems return.
Only if you already take them, through your existing payment provider. Card details never pass through the agent.
It acknowledges the complaint, stops, and passes it to you with the booking attached. It does not offer refunds or explanations.
About a week: table rules, sittings, the allergen routing list and a parallel run over two services.
Start here
Show us a week of bookings.
One hour with your table plan and last week's booking log. We will show you how many calls went unanswered and what the agent would have taken.