After the weekend, you want to improve the menu and check which items were ordered most often. You have recipes, current prices and an order summary. AI—a program that can generate text and identify patterns in supplied data—can prepare a first draft or organise a table. It does not know your kitchen, supplier agreements or what happened during service.

1. Start with a specific question

“Improve my menu” is too broad. A better task is: “Shorten this description to two sentences, keep every ingredient and add nothing” or “Group these anonymous order counts by time of day and flag items for further review.” The output is then easier to assess.

Before you start, define what the program will receive and what it should return. A text task may need only an approved recipe, preparation method and target length. A simple order summary needs item names, order counts, a date range and sales channel. Removing unnecessary information makes the process easier to control.

2. Draft menu descriptions and translations

AI can suggest a shorter category name, a first draft of a dish description or a working translation. Supply the current ingredients and state what must not change. Then compare the text with the recipe and ask a fluent speaker to review the translation.

Keep a short review list: ingredients, preparation method, serving size, price, variants and language. Revisit the description when the recipe changes. A previously correct version may no longer match the plate.

If you only need help with copy, read the guide on how AI can help write dish descriptions. This article goes further by covering order data and a repeatable review process.

3. Read aggregated order data

An order summary can help the team find questions worth investigating. For example, an item may be popular for delivery but rarely ordered in the dining room. That alone is not a reason to change the price or remove the dish. Check whether both channels used the same offer, availability hours and comparison period.

Prepare an aggregated table without names, phone numbers, addresses or notes from identifiable guests. Ask the tool to organise the data, identify missing values and show how it calculated its result. Then compare the answer with the source table. NIST presents generative AI risk management as work across the full lifecycle; for a restaurant, that includes retaining the input and being able to verify the output.

A conclusion should lead to a test, not an automatic decision. You might improve the visibility of an overlooked category for one week, then compare similar days. Do not credit one change when weather, promotions, availability or opening hours changed at the same time.

4. Keep important decisions with the team

Do not delegate allergen decisions, live availability, prices, refunds or complaint responses to the program. AI may produce a draft or a list of questions, but an accountable employee must check the facts and approve publication.

Take care with personal data too. The European Data Protection Board says that the use of personal data in developing and deploying AI models requires case-by-case assessment under GDPR principles. Before sending data to an external tool, check the service terms, processing purpose and your organisation’s rules. Aggregated data that cannot identify a guest is the safer place to start.

AI prepares

A working description, category suggestion, table summary, list of missing information and questions for further review.

The team decides

What matches the recipe, which price applies, what is available, which data may be used and what guests finally see.

5. Introduce a simple review routine

Start with thirty minutes once a week. Choose one menu category or one order report. Save the source version, the instruction and the accepted change. After a few weeks, you can see whether the tool reduces work or merely creates another draft to correct.

Checklist before publishing or deciding

  • Are the recipe, price and reporting period current?
  • Have you removed data that could identify a guest?
  • Did the program invent an ingredient, feature or explanation?
  • Did the responsible employee compare the result with the source?
  • Is it clear who approves the change and how it can be reversed?

To connect a QR menu with ordering, explore the iTasty online ordering page. During a demo, ask which AI assistant capabilities are available, what data a chosen process uses and what it costs. Do not assume that the applications described here are included automatically in every iTasty plan.

Sources

Sources checked on 13 October 2026.