AI automation for food service with the numbers the same day
This is written for the owner of several restaurants or a small chain, not for a two-table bistro. With one site you see everything from the floor. With five, every report arrives in a different format and a day late, so you drive between sites to find out what happened. We pull sales, stock and rota data into one report that is waiting for you in the morning.
- 14 daysto the first automation running, counted from the end of the workshop
- EUR 3,500net per week of workshop, a published price we do not negotiate
- 5 to 10 secondsfor the AI agent to answer a manager out of your own recipes and procedures
Problem
Why does the owner of five sites learn the result a month late?
Every site reports its own way. One manager sends the till total by message in the evening, the second types it into a spreadsheet the next morning, the third remembers he meant to. By the time the numbers line up, they are out of date. Food cost gets calculated when the quarter closes, which is exactly when it is too late to change a price on the menu.
The rest of the day looks the same. Supplier orders go out from the manager's memory, on a messenger. The rota lives in a spreadsheet and changes on Friday evening. Bookings and guest questions come in by phone, by form and from the map listing, each landing somewhere else. So the owner drives between sites, because that is the only reliable way to see what is actually going on.
The rota is the same story, except here the cost lands on people rather than on margin. Shifts have to be published far enough ahead for a cook to plan a week around them, and they have to hold for the whole period instead of moving every Friday evening. A spreadsheet patched on the fly will not carry that, especially with the turnover this industry runs on. Every swap then travels by phone, and nobody except the manager knows which version is current.
- Yesterday's takings assembled from five messages, each in a different format
- Food cost calculated once a quarter, so the menu price gets fixed after the season
- The supplier order written from the manager's memory, on a messenger
- Bookings and guest questions in three channels, with no shared calendar
- You know how a site did only once you have driven to it
Why now
What changed that makes this pay off today
Point-of-sale systems on the floor and delivery platforms now expose interfaces you can read sales through without exporting to a spreadsheet. Wiring five sites into one report is counted in working days rather than in quarters, and nothing behind the bar gets replaced. The automation layer is documented in public: n8n and Make both publish what their connectors do, so nobody is guessing.
The second change matters more. A rule will carry the till total from the point of sale into a spreadsheet, but it will not tell a manager how many grams of cheese go on the pizza under the new recipe. An AI agent answers out of your own documents in 5 to 10 seconds, so the question stops landing on your phone in the middle of service.
The third one is technical and concerns the software on the office computer. Cloud invoicing tools we read through their web interface. An accounting program installed locally we reach through a connector running inside your own network, because the program never leaves the building. Both routes are standard practice today, not a research project.
What changes
What do you see after the first month?
One report from every site, every morning
Sales, costs, food cost and takings per site, pulled from the point of sale, the stock system and the delivery platforms into a single table. It is waiting at a fixed hour, before you open the first site.
The same format for every site, including the one opened last month
An alert when food cost or takings drift
You set the threshold per site and per category. The message goes to Slack or to email the same day, instead of surfacing at the monthly review.
A deviation is visible on the day it happens
Supplier orders from stock, not from memory
The system works out consumption from recipes and sales, compares it against stock on hand and the plan for the coming days. The manager approves a finished list instead of typing it out on a messenger at closing time.
Separate thresholds for fresh and dry goods, with supplier lead times
Bookings and guest questions in one place
Phone, form, map listing and messages land in one inbox, with a confirmation and a reminder before the visit. The manager keeps one calendar rather than three.
One calendar per site instead of three unsynced channels
Managers ask the AI agent, not you
Recipes, portion weights, plating standards and the opening and closing checklists sit in one place, and the agent answers out of them in a plain sentence. An answer in 5 to 10 seconds, from a document you approved.
An answer from your own documentation, not from the internet
Checklists and documents in one flow
Opening and closing checklists, records and HACCP paperwork go into a shared archive instead of a binder in each site. You can see which site has not filled in what.
A missing checklist shows up the same day, not before an inspection
Proof
Do you use this yourselves?
This is not a rollout in a restaurant group and we are not going to pretend it is. These are the only figures we may show without asking a client for permission to publish their data: an AI agent running in production at our founders' agency since March 2026, built by the same method described above. We give them together with what they do NOT prove.
Built for ourselves. These are not client deployments but implementations inside the founders' own companies. We say so plainly, because proof you cannot verify is not proof.
JustBottheir agentJustIdea Agency, a performance marketing agency of around 30 people
- 23
- people from the company work with the agent
- 38
- systems wired in permanently
- 112
- processes run without a human
- 475
- thousand lines of the agent’s code
An agent wired into Google Ads, Meta, the CRM, the task system, the SEO tool, Analytics, Search Console and the data warehouse. First commit 1 March 2026, today 23 people use it.
8 of 38 systems it works in
Google Ads
Meta Ads- HubSpot
- ActiveCollab
- Ahrefs
GA4
Search Console
BigQuery
- Before
- In the twelve months before launch the agency ran at a margin of minus 4.42 percent. Across a twenty-month window there was not a single run of three profitable months.
- After
- May, June and July 2026 are three profitable months in a row, the first such run in the whole window. Revenue barely moved, so the change sits on the cost of delivery, not on sales.
Margin up 5.55 percentage points year on year, with revenue up 1.5 percent
What about us
What if our setup is different?
Four sentences come up in almost every restaurant. Here are the answers we would give you on a call.
We have our own point-of-sale system and we are not changing it.
No need to. We read data out of it and leave it where it stands. Swapping the point of sale across several sites is a project for a whole season, and this is about 14 days to the first result.
Automation will not make the food better.
It will not, and we do not promise that. It replaces neither the chef nor the waiter. It takes reporting and ordering off your people and puts the numbers in front of the owner the same day instead of a month later. The plate is still the kitchen's job and the room is still the floor team's.
We run three sites, that is too small for this to pay off.
At three sites the return sits in reporting and in ordering, not in advanced analytics. At fifteen you add a data warehouse and site-against-site comparison. At the workshop we say which of those is real for you, and if neither is, we say that too.
Our managers will not type data into yet another system.
That is why we start from the data that already creates itself: sales from the point of sale, orders from the delivery platforms, invoices from suppliers. Manual entry goes in only where the process will not close without it, and in the simplest form we can find.
- Sales on the floor
- point-of-sale systems, delivery platforms, payment terminals
- Stock and purchasing
- stock levels, recipes, supplier orders, purchase invoices
- Accounting
- cloud invoicing tools, locally installed accounting software, supplier invoice inboxes
- Team work
- Google Workspace, Microsoft 365, Slack, rota spreadsheets
- Data and reporting
- Google Sheets, BigQuery once the group passes a dozen sites
- Automation layer
- Make, n8n, custom integrations over an API
What exactly we automate in a restaurant group
A typical split of the work in food service before the rollout and after it.
| Area | Today | After the rollout |
|---|---|---|
| Daily report from the sites | each manager sends it his own way, in the evening or the next day | one report from every site at a fixed hour in the morning |
| Food cost | calculated when the quarter closes | calculated from recipes and sales, with an alert on a deviation |
| Supplier orders | from the manager's memory, on a messenger | a list built from stock and the plan, the manager approves it |
| Rota | a spreadsheet patched on Friday evening | published ahead of the period, every change in one place |
| Bookings and guest questions | phone, form and map listing kept apart | one calendar and one inbox per site |
| Procedures and recipes | a binder in each site, versions drift apart | one source, the AI agent answers managers out of it |
| HACCP paperwork | paper in the site, hunted down before an inspection | documents and checklists in one archive |
The scope of the first rollout gets set at the workshop. We take the three items with the highest return, not all of them at once.
How we work
Five steps from five reports to one
Every step ends with something you can judge before you decide on the next one.
Workshop
One week. We walk a day in the site from delivery through service to cashing up, and we follow one number from the receipt to the report that reaches you. We come out with a map and a list of the places where the same number gets produced twice.
Picking three automations
From that map we take the three with the highest return and put numbers on them before we start, so that after 30 days there is a way to tell whether they worked.
Building the first one
14 days to something running. In food service we usually start with the single report covering every site, because that is the one that ends the morning round of phone calls to managers.
Pilot in one site
The first version runs in one site, usually the one with the weakest reporting. We test it on real sales and real stock before it goes anywhere else.
The rest of the sites and the measurement
We roll it out site by site, each one on the same format. After 30 days and after 90 days we check the indicators agreed at the workshop, and only then decide on the next area.
Before and after
How the week of a group owner changes
Before
- Rings round the managers in the morning for yesterday's takings
- Learns the food cost a quarter later, when the menu prices are already stale
- Drives between sites to see what is going on
- Approves the supplier order from memory, over the phone
After
- The report from every site is waiting in the morning, ready to be discussed
- A food cost deviation reports itself, on the day it happens
- Drives to a site to talk to the team, not to collect numbers
- The order is calculated from stock and the manager approves it
The number of sites to watch does not change. What changes is whether you hear about a problem the same day or a month later.
For the owner and the operations director
What the person signing this off gets out of it
The first thing shows up immediately: you stop starting the day by collecting numbers and start it by making decisions. The morning calls to managers disappear, because the report is already there, and the conversation is about what to do with the figure rather than what the figure is.
The second is quieter, and it matters more when you open the next site or sit down with a bank. The profitability of one site and of one menu item stops being knowledge held by one person and becomes a number in a system that you can put on a screen. A new site joins a reporting format that already exists instead of inventing its own.
- The same report format in every site, including the one opened last week
- Food cost and takings deviations flagged on the day they happen
- Recipes and procedures written down in one place, so the manager does not call you
- The rollout goes site by site, so no service stops because of it
What it costs
The workshop price is published, the rollout price comes after it
The workshop costs EUR 3,500 net per week and we publish that before the first call, because a company has a right to know what it is agreeing to. One site usually takes a week, a group of several takes two or three. The rollout gets quoted after the workshop, once we know how many systems we are connecting and what condition they are in.
Tool and model costs you pay straight to the provider, with no commission from us. A single model query runs from EUR 0.005 to 0.07 depending on the model, and that invoice comes to you anyway, so we do not sit in the middle of it.
Process workshop
EUR 3,500 net
per week, usually one to three weeks for a restaurant group
A closed piece of work. It ends with a process map and a 30-day plan you could carry out without us.
- A week of work in the sites and in the systems
- A map of the route a number takes from the receipt to the owner's report
- The three automations with the highest return, costed
- Indicators to measure after 30 and 90 days
- Most popular
Workshop and first rollout
quoted after the workshop
the scope follows from the number of sites and systems
For a group that wants one report covering every site within two weeks of the workshop ending.
- Everything in the workshop
- The first automation built in 14 days
- A connection to the point of sale, the stock system and the delivery platforms
- A pilot in one site, then the others
- Training for the person who runs this on your side
AI agent for a restaurant group
quoted after the workshop
for groups where managers ask the same questions every day
Everything in the rollout plus BOND, the company AI agent that answers out of your procedures and your data.
- Everything in the rollout
- An agent answering on sales, stock, food cost and recipes
- Alerts on food cost and takings deviations per site
- Access for site managers, the kitchen and the back office
If after the workshop we conclude there is nothing worth automating, we say so plainly and do not propose a rollout. The workshop still ends with a document that stands on its own: a process map and a 30-day plan you can carry out yourself or with anyone else.
Frequently asked questions
What does automation mean in food service?
It means a system taking over work somebody does by hand today: collecting takings from the sites, calculating food cost, building supplier orders and answering the same manager questions over and over. The data for that usually exists already, it just sits in the point of sale, the stock system, the rota spreadsheet and the delivery platforms at the same time. Automation is about pulling it into one place, not about buying another program.
Do we have to replace the point-of-sale system?
No. We read data out of it and leave it where it stands. Replacing the point of sale across several sites is a project for a whole season, and the first working automation is meant to exist 14 days after the workshop ends. The same goes for the delivery platforms: we pull sales and commissions out of them and change nothing inside them.
How do you calculate food cost?
From recipes and from sales: how many items went out, how much raw material that consumed, at what purchase price from the most recent invoices. We line that up against takings per site and per menu category. Frequency is the point here, because a food cost calculated once a quarter tells you about something you can no longer react to in the same season.
Will this work with three sites?
It will, but the return sits somewhere other than at fifteen sites. With three, the gain comes from the shared report and from orders calculated off stock. In a larger group you add a data warehouse such as <a href="https://cloud.google.com/bigquery/docs" rel="noopener">BigQuery</a> and start comparing sites against each other. At the workshop we say which of those is worth the money in your case.
What about delivery platform commissions?
We pull sales and the commissions charged along with the rest of the data, so margin after commission sits next to margin from the dining room. The level of the commission is a matter of your contract with the platform and we do not change that. What changes is that you can see what the channel costs you in a given site before you close the month.
Will the AI agent answer a manager out of our procedures?
Yes, and in food service that is its main use. BOND, our company AI agent, answers out of the documents you point it at: recipes, portion weights, plating standards, opening and closing checklists. The answer comes back in 5 to 10 seconds. If something is not in your documents, the agent says it does not know instead of guessing.
How long does a rollout in a restaurant group take?
A week of workshop for a single site, two or three for a group, then 14 days for the first automation. A full AI agent takes around two months. Extending to further sites goes faster than the first rollout, because the report format and the integrations already exist by then.
What do you do with HACCP paperwork?
We treat it as a flow of documents and checklists, not as food safety consulting. Opening and closing checklists, records and measurements filled in at the site go into one archive with a date and a name against them, instead of into a binder. You can see which site has not completed what, the same day. Interpreting the rules stays with your own specialist.
Let us start with one site
The free consultation takes 30 minutes. We walk one of your processes, usually the route a number takes from the receipt to the report you get in the morning, and we tell you whether we see anything there worth automating. If we do not, we say so on the call. You do not have to give us access to any system at this stage.
We run at most two rollouts at a time, so that somebody is actually sitting on each of them.
Book a free consultationPS. If you spent this morning ringing round managers for yesterday's takings, this is exactly the problem. Automation will not improve the food and it will not replace a waiter. It will take reporting and ordering off your people and put the numbers from every site on your desk the same day.