You run a business that lives or dies on the next six hours, and the office work never fits. A chat assistant drafts the schedule from your availability sheet, answers the two-star review in a tone you would sign, turns a POS export into a food-cost report, and writes the allergen quiz for pre-shift. It does the paperwork between the lunch rush and the dinner push.
It does not know your kitchen. It will describe a cooling procedure that violates the Food Code, invent an allergen fact that could hurt someone, botch a recipe scale-up, and write a schedule that breaks your state's break or minor-labor rules. It cannot smell the walk-in. Anything that touches food safety, fire and life safety, or wages gets checked against the actual rule by a person.
Guests and staff trust you with their data: reservations, card numbers, allergies, employee Social Security numbers, tip records. Keep all of it out of consumer AI tools unless your company has approved the tool for that use, and keep review responses honest and specific, because guests can tell when a template is answering them.
Quick wins this week
- Paste last month's Google and Yelp reviews with guest names removed and ask for the three recurring complaints and compliments, then answer the worst one with a draft you rewrite in your own words.
- Give the AI your menu and ask for descriptions that sell without adjectives you would never say out loud, then cut them to two lines each.
- Turn your opening and closing procedures into a printed checklist with a line-cook version and a front-of-house version.
- Paste a POS product-mix export and ask for the five items with the worst margin-to-popularity ratio, then verify the math on your top seller by hand.
What AI can do for restaurant managers, task by task
Schedule drafts from availability and forecast
Paste availability, your sales forecast by daypart, positions needed per shift, and the rules your HR or payroll provider has given you (max hours, break timing, minor restrictions, advance-notice rules), and ask for a draft with gaps and overtime risk flagged. It is a starting point for your scheduling app, not a posted schedule. Every draft is checked against state and local labor law before it goes out.
Responding to online reviews
Paste the review with the guest's name removed, describe what actually happened and what you did about it, and ask for a short, specific reply with no excuses. Rewrite it so it sounds like you. Never confirm details of a guest's visit publicly, never argue, and take anything involving illness, injury, or a legal threat to your owner or insurer before replying.
Menu descriptions, specials, and promotions
Give the dish, the ingredients, the method, the price point, and your restaurant's personality, and ask for two-line descriptions and a specials board. Verify every ingredient and allergen statement against the recipe card; the model will invent a 'gluten-free' or 'nut-free' claim. Marketing copy about allergens is a safety document.
Food cost, product mix, and inventory analysis
Export product mix and inventory counts, remove any employee or guest data, describe the columns, and ask for food cost percentage by category, theoretical versus actual variance, and the menu-engineering quadrant for each item. Ask for the calculations and recompute your top five items by hand. Models make arithmetic errors and confuse portion units.
SOPs, checklists, and line-level training
Describe how a station actually runs and ask for a numbered procedure, a laminated checklist version, and a short quiz, in the languages your crew speaks with a bilingual staff member checking the translation. For any food safety step (cooking temperatures, cooling, holding, date labeling, allergen handling), verify against the FDA Food Code as adopted by your health department before it goes on the wall.
Recipe scaling, prep lists, and ordering
Paste the recipe with yields and ask for a scaled version, a prep list by station, and order quantities based on the forecast you supply. Check the math on every scaled ingredient and every conversion; models get ounces, grams, and volume-to-weight wrong. The order still gets a human look against what is actually in the walk-in.
Prompts for restaurant managers
Replace the bracketed placeholders, paste into any chat assistant, and iterate on the result.
Schedule draft with labor rules flagged
Act as a restaurant scheduler. Build a draft schedule for the week of [DATES] for a [RESTAURANT TYPE] with these positions needed per shift: [POSITIONS BY SHIFT AND DAY]. Staff availability and max hours: [PASTE AVAILABILITY]. Sales forecast by daypart: [FORECAST]. Rules I must follow: [BREAK RULES, MAX HOURS, MINOR RESTRICTIONS, ADVANCE NOTICE RULES, TIP POOL ROLES]. Output a table by day and shift with names, then a list of every gap, every shift that would push someone into overtime, and every place a rule I gave you would be broken. List the assumptions you made. This is a draft for review, not a schedule to post.
Tip: Your payroll provider or HR resource confirms the rules for your state and city. Predictive scheduling laws apply in some places and change.
Review response you would actually sign
You are the general manager of [RESTAURANT NAME AND STYLE]. Reply to this online review, with the guest's name removed: [PASTE REVIEW]. What actually happened, as best we know: [YOUR ACCOUNT]. What we have done or will do about it: [ACTION]. Under 90 words, warm, specific, no excuses, no restating the complaint, no confirming details of their visit beyond what they wrote, no offers of free food in public. Thank them, own what we own, say what changed, and invite them to reach me directly by [CONTACT METHOD]. Sound like a person, not a brand. Then give me a version for a positive review that mentions [SPECIFIC DETAIL THEY PRAISED].
Tip: Reviews mentioning illness, injury, or a legal threat go to your owner or insurer before you reply. Do not let the model answer those.
Menu-engineering analysis from a POS export
You are a restaurant financial analyst. Here is a product-mix export with no guest or employee data. Columns: [DESCRIBE COLUMNS, INCLUDING ITEMS SOLD, PRICE, AND PLATE COST]. Data: [PASTE DATA OR ATTACH SPREADSHEET]. Period: [DATES]. Calculate contribution margin per item, sort items into menu-engineering quadrants (stars, plowhorses, puzzles, dogs) using average popularity and average margin as the cutoffs, and show your calculations. Then recommend three actions: one price change, one menu placement change, one item to rework or cut, each with the numbers. Flag any item where the plate cost looks wrong or missing.
Tip: Recompute the margin on your top five sellers by hand. One bad plate cost in the export skews the whole quadrant.
Food safety SOP with verification flags
Act as a restaurant trainer writing a station SOP for [TASK OR STATION] in a [RESTAURANT TYPE]. Here is how we do it: [PASTE NOTES]. Equipment: [EQUIPMENT]. Languages spoken by staff: [LANGUAGES]. Format: purpose, setup, numbered steps, a laminated checklist version, and a five-question quiz with answers. Mark every step involving cooking temperatures, cooling, hot or cold holding, date labeling, allergen handling, or sanitizer concentration with 'VERIFY AGAINST FOOD CODE' and do not state specific temperatures or times; I will add those from the FDA Food Code and our local health department requirements. Then produce the checklist in [SECOND LANGUAGE].
Tip: Temperatures and times come from the Food Code and your health department, never from the model. Have a bilingual staff member check the translation.
Recipe scale-up with double-checked math
You are a chef who double-checks arithmetic. Scale this recipe from [ORIGINAL YIELD] to [TARGET YIELD]: [PASTE RECIPE WITH UNITS]. Keep the original units unless converting to weight makes the recipe more accurate, and if you convert, show the conversion factor. Output the scaled recipe, a prep list by station, and an order quantity for each ingredient given a forecast of [NUMBER OF PORTIONS] over [PERIOD] and current on-hand of [ON-HAND QUANTITIES]. Then recompute every scaled amount independently and flag any ingredient that does not scale linearly (leavening, salt, spices, thickeners) with a note on what to adjust.
Tip: Check every conversion. Volume-to-weight is where the model fails, and the walk-in still gets a human look before you order.
Want a prompt for something else? Use the Prompt Builder.
Skills to build
Keeping food safety, fire safety, and wages off the model's plate
Why: Cooking and cooling temperatures, allergen handling, occupancy and exit rules, hood suppression, and wage-and-hour compliance have real rulebooks and real consequences. A confident answer from a chat assistant is not a rule.
How: Post a short list of decisions that always go to the Food Code, the fire marshal, or your HR resource. Use AI to format and teach those rules, never to set them.
Prompting with your real numbers and rules
Why: The model does not know your labor target, your plate costs, or your state's break rules. Useful output depends on what you paste in.
How: Keep a one-paragraph restaurant profile (concept, dayparts, headcount, labor and food cost targets, the labor rules HR gave you) and paste it at the top of every operational prompt.
Protecting guest and employee data by reflex
Why: Reservation records, card numbers, allergy notes tied to names, employee Social Security numbers, and tip records are confidential, and card data is covered by PCI rules. A paste into a consumer tool is a data exposure.
How: Strip names, card details, and employee data before any paste. Keep those in your POS, reservation, and payroll systems. Ask your owner or company which tools are approved.
Reading generated numbers skeptically
Why: Food cost, scaled recipes, and labor percentages are arithmetic, and models get arithmetic wrong while sounding sure. A wrong plate cost changes a menu decision.
How: Recompute the top five items or the biggest line by hand every time. Ask the model to show its calculations so you can check them.
Answering reviews like a person
Why: Guests recognize templates and punish them. A specific, honest reply to a bad review wins back the reviewer and reassures everyone reading it.
How: Use the model for a first draft, then rewrite the first sentence in your own words and add one detail only you would know. Never let it answer illness, injury, or legal threats.
Tools worth knowing
ChatGPT
The general-purpose AI assistant most of your coworkers already use.
Gemini
Google's assistant, strongest when your work already lives in Google Workspace.
Microsoft Copilot
AI inside Word, Excel, Outlook, and Teams, with your company's data protections.
Claude
A careful writing and analysis assistant that shines on long documents.
Canva
Design for non-designers, with Magic Studio AI for text, images, video, and resizing.
Zapier
Connect thousands of apps with no-code workflows, now with AI steps, agents, and chatbots
Cautions for restaurant managers
Cooking, cooling, and holding temperatures, date labeling, sanitizer concentrations, and allergen handling are governed by the FDA Food Code as adopted by your state and local health department, and by your food safety certification training. Models state wrong temperatures and invent allergen claims. Verify every food safety step and every allergen statement against the actual rule and the recipe card before it reaches the wall, the menu, or a guest.
Occupancy limits, exit access, hood and suppression system maintenance, and emergency procedures come from your fire marshal and your inspection reports. Use AI to write the training and the checklist, never to decide what the rule is or whether you are in compliance.
Tip credit and tip pooling rules, overtime, meal and rest breaks, minor work hours, and predictive scheduling ordinances vary by state and city and change often. A schedule or pay decision the model drafts is not compliant until your HR resource or payroll provider says it is. Do not ask a model to decide who is in the tip pool or how to handle final pay.
Reservations, guest allergy notes, card numbers, loyalty data, employee Social Security numbers, I-9 documents, and tip records are confidential, and card data is covered by PCI DSS. Never paste confidential guest or employee data into a consumer AI tool unless your company has approved that tool for that use. Keep it in the systems built to hold it.
Never confirm a guest's visit details publicly, never argue, and never let a model reply to a review that mentions illness, injury, discrimination, or a legal claim; those go to your owner or insurer first. Templated replies are easy to spot and cost you more than silence.
Your 30-day plan
- Week 1: Ask your owner or company which AI tools are approved and what may be pasted. Write your restaurant profile paragraph. Use AI for one low-stakes task a day: a specials description, a shift announcement, a vendor email.
- Week 2: Run the review-response prompt on your last five reviews and post the ones you have rewritten. Turn one station's procedure into an SOP and checklist, verify the food safety steps against the Food Code, and put it on the wall.
- Week 3: Export one month of product mix and run the menu-engineering prompt. Recompute your top five items by hand and make one menu change based on what holds up.
- Week 4: Draft next week's schedule from the prompt, review it with your HR resource against your state's rules, and compare the time to your usual process. Build one allergen quiz for pre-shift from your own written procedure.
- End of month: Pick the three tasks you will keep using AI for, start a shared folder of prompts and SOPs, and write down the mistakes you caught (a wrong temperature, a bad conversion, an invented allergen claim) to share with your managers.
Frequently asked questions
Can AI make a restaurant schedule?
Should restaurants use AI to respond to reviews?
Can AI help with food cost and inventory?
Is it safe to put employee or guest information into ChatGPT?
Can AI write food safety procedures for my restaurant?
Terms used on this page
Related roles
- AI for Hotel ManagersUse AI to get guest correspondence, review responses, SOPs, staffing drafts, and owner reports done faster, while guest data stays in your PMS and fire, life safety, and labor decisions stay with the people accountable for them.
- AI for Retail ManagersStore managers have no desk time, which is why AI matters: the recap, the schedule draft, the review reply and the new-hire guide can all be done from your phone in the back room. Here is how, and what never goes into a chat tool.
- AI for Small Business OwnersAI answers the one-star review calmly, drafts the month of posts, turns your voice memo into an SOP, and preps the questions for your CPA. You still make the calls, sign the checks, and own what goes out under your name.
- AI for Event PlannersEvent planning is a thousand details and a hundred emails a day. AI is good at the details that are writing, lists and comparisons: run-of-show, RFPs, attendee messages, recaps. Here is how to use it, and why contracts and attendee data stay in your hands.