Analyses on our own data
Not one article called “5 ways to cut food cost”. We write only about what we see in the invoice flow of real restaurants - with the numbers and the method behind them.
How long it takes to find out one number
A simple question like «how much did we overpay for meat in August» travels through the mailbox, the accounting system, an export and a spreadsheet. It takes from half an hour to half a day, and the answer arrives after the decision was made without it. Not a matter of skill: the number is assembled by hand.
Margin is counted once a month and lost every day
A report at month end shows the result when nothing can be done about it: prices rose on delivery day, write-offs happened on a shift, an extra person went on the rota three weeks ago. Margin is built from five floors, and on each of them a decision only means something on the day of the event.
Digitising a restaurant without replacing systems: the till and the books stay
Replacing the accounting system costs months of work and retraining, and does not solve the original problem: scattered data. It is cheaper to leave the till, the books and the rota where they are and add a layer that reads them, brings the data to one shape and checks it. The system of record stays the one you already run.
Seven tasks an AI agent closes in a restaurant today
An agent closes tasks that have an input, a rule and a checkable result: receiving documents, matching them to the catalogue, arithmetic checks, price control, dish cost, preparing the reconciliation and answering questions from the data. Decisions about suppliers, menu prices and people stay with the team - the agent prepares them, it does not take them.
Why AI cannot answer a question about your restaurant
Because the data sits in separate systems: sales in the till, purchasing in the books, shifts in the rota, invoices in the mailbox, and some figures only in the manager's head. None of these systems can see the others, so any question that crosses two of them goes unanswered.
MCP: how to connect Claude or ChatGPT to your restaurant's data
MCP is an open protocol through which external AI reaches data by a set of read tools rather than by file exports. The owner approves the connection, the AI requests the slices it needs itself, every call is logged, nothing is written back into the restaurant's systems, and access is revoked in one action.
Till, books, accountant: why every one of them has a different number
Each source counts a different thing: the till counts rung-up receipts, the books count posted documents, the accountant counts documents with paper originals and VAT. A difference between them is not an error in itself but a consequence of different period boundaries, rounding and missing documents. The question is not who is right but what exactly differs.
AI in restaurants: what already works and what is still a demo
AI works in a restaurant wherever the data is already collected and normalised: reading supplier documents, checking purchase prices, recalculating dish cost, answering questions about your own figures. Where data sits in separate systems and was never reconciled, AI stays a demo and answers in generalities instead of talking about your restaurant.