26% of restaurant operators report using AI-related tools, with marketing the top use — 19% of full-service and 15% of limited-service operators — per the National Restaurant Association's 2026 State of the Restaurant Industry report.Source ↗
In Toast's 2025 Voice of the Restaurant Industry Survey of 712 US restaurant decision-makers, 86% said they were comfortable using AI and 81% planned to use it more.Source ↗
Restaurant managers spend an estimated 11-15 hours a week on administrative work, about 13% of it on scheduling, according to a Paychex survey cited by FSR Magazine.Source ↗
The US Department of Labor reaffirmed in December 2024 (opinion letter FLSA2024-02) that managers and supervisors may not keep or share in employees' tips through a tip pool.Source ↗
planningChatGPTClaude

Turning the week's forecast into a draft schedule and shift notes

Every week starts with the same puzzle — fit everyone's availability and time-off requests against forecasted covers, then write the shift notes that tell each crew what to expect. It is the single biggest administrative task most managers carry, and the one that eats a night off. AI can assemble the first draft in seconds, as long as the manager keeps the legal calls.

Prompt
You are a shift-planning assistant for a restaurant manager. Build a DRAFT weekly schedule and shift notes from the information below. You are not making final labor decisions — I am.

Forecasted covers by day and daypart: {{forecast_by_daypart}}
Staff, roles, and availability (initials only): {{staff_availability}}
My scheduling rules (min/max hours, who can open or close, required breaks, events): {{scheduling_rules}}

Produce:
1. A draft schedule grid by day and station, using ONLY the people and hours I listed.
2. A 2-3 line shift note per day covering events, sections, and anyone training.
3. A coverage-gap list: any shift with no available person marked [NEEDS COVERAGE].

Rules: never assign anyone outside their stated availability. Do not calculate overtime, predictability pay, or any tip split — flag those as [VERIFY WITH MANAGER/PAYROLL]. Use only the initials I provided; invent no staff and no availability. If my rules conflict (more coverage needed than people available), say so plainly instead of guessing.

Fill in your details and the prompt updates live — then copy.

communicationChatGPTClaudeGemini

Replying to Google and Yelp reviews without sounding canned

Diners read the reviews before they ever call, and one unanswered one-star sits at the top of the profile doing damage every day. Managers now use AI to draft replies quickly, but a public response has to stay truthful, protect the guest's privacy, and never read like a corporate template.

Prompt
You are the reputation manager for a restaurant. Draft public responses to the review below.

The review, pasted exactly: {{review_text}}
What actually happened, for your context only — do NOT reveal any of this in the reply: {{internal_context}}
Our voice: {{brand_voice}}

Write three options — brief (under 40 words), standard (under 80), and longer (under 120). Every option must:
- Acknowledge the experience without admitting legal fault.
- Never reveal the guest's order, table, or account, and never confirm private details.
- Offer a specific offline path (name plus phone or email) to make it right.
- Sound like a real person — no "we take this seriously" boilerplate, no sarcasm.
- Say ONLY what I have confirmed above. Do not invent an apology story, a cause, or a refund I have not approved.

STOP condition: if the review alleges food poisoning, injury, discrimination, or threatens legal action, do not draft a reply. Output only "ESCALATE: route to owner and insurer before responding" and list the allegations you spotted.

Fill in your details and the prompt updates live — then copy.

writingClaudeChatGPT

Drafting station SOPs and training checklists new hires can follow

Turnover never stops, and consistency across shifts lives or dies on written procedures — but the SOPs and training checklists are exactly the documents that never get finished. AI can turn a manager's rough notes or a quick walkthrough into a clean, step-by-step SOP in minutes, as long as the food-safety specifics come from a human.

Prompt
You are an operations trainer writing a restaurant SOP and a training checklist. Build them from MY steps only.

The station or task: {{task_or_station}}
How I actually do it (rough notes or a transcript): {{my_steps}}
Equipment and tools involved: {{equipment_and_tools}}

Produce:
1. A titled SOP with purpose, scope, step-by-step procedure, and who is responsible.
2. A one-page checklist a new hire can tick off on their first shifts.
3. Three short quiz questions to confirm they understood the steps.

Rules: use ONLY the steps and rules I gave you. Do NOT invent cooking or holding temperatures, cooling times, hold times, allergen procedures, or health-code specifics. Anything safety-related that I did not state, mark [CONFIRM WITH FOOD SAFETY / LOCAL HEALTH CODE] rather than filling in a number. Keep the language plain and short — this is for someone on day one.

Fill in your details and the prompt updates live — then copy.

creativeChatGPTGeminiClaude

A month of local promos, captions, and menu copy in one sitting

Marketing is the first thing that slips when the floor gets busy, yet slow Tuesdays and seasonal specials need a steady drip of posts and emails. Marketing is the single most common AI use in restaurants, and a manager can batch a month of content in one sitting instead of scrambling the morning of.

Prompt
You are a marketing assistant for an independent restaurant. Build a month of local promotional content from the details below.

The offer or theme: {{offer_or_theme}}
Cuisine and neighborhood: {{cuisine_and_location}}
Brand voice: {{brand_voice}}
Channels and key dates: {{channels_and_dates}}

Produce:
1. A 4-week content calendar mapping posts to dates.
2. Instagram captions for each post, plus a matching set of hashtags.
3. Two email subject lines and one short email body.
4. One SMS blast under 160 characters.
5. Three menu-item descriptions, appetizing but honest.

Rules: use ONLY the offers, prices, dates, and dishes I gave you. Do NOT invent awards, "best in town" claims, health or nutrition benefits, or ingredients. Keep every promo inside the terms I set. Flag anything that needs a dietary or legal check as [VERIFY]. If a date or price is missing, ask rather than guess.

Fill in your details and the prompt updates live — then copy.

analysisChatGPTClaude

Turning the weekly POS export into a plain-English shift report

The POS produces a wall of numbers; what a manager and staff actually want is the story — what sold, what is slipping, and where labor or food cost ran hot. Instead of building a spreadsheet, managers paste the export and ask for a short, plain summary they can act on and share.

Prompt
You are a restaurant analyst. Turn the numbers below into a short, plain-English weekly summary.

This week's POS and labor figures, with last week's for comparison: {{pos_numbers}}
Anything unusual that week (weather, event, closure): {{context_notes}}

Write:
1. A 150-word summary. Bottom line first (net sales and the single biggest change from last week). Then the largest driver explained in plain terms. Then three items trending up and three trending down. Close with one thing to test next week.
2. A three-bullet "at a glance" block: net sales, labor %, food cost %.

Rules: use ONLY the numbers I paste. Always name the comparison ("down 8% vs. last week"). If a figure I referenced is missing, write [MISSING] rather than estimating it. Do not forecast future revenue or make any guarantee. Do not recompute percentages I already provided — quote mine.

Fill in your details and the prompt updates live — then copy.

automationChatGPTClaude

Restock reminders and supplier emails from your usage patterns

Ordering runs on par levels and a steady back-and-forth with vendors — reorder emails, pushback on a price increase, a complaint about a short or wrong delivery, a substitution request. Managers use AI to draft these fast from their on-hand notes, keeping the tone professional even at the end of a double.

Prompt
You are a purchasing assistant for a restaurant manager. Help me order and write to a vendor.

Items with current on-hand vs. par: {{inventory_and_pars}}
Next week's forecast or events: {{forecast_or_events}}
The vendor and the issue: {{vendor_and_issue}}

Produce:
1. A suggested order quantity per item, each with a one-line reason tied to par, on-hand, and the forecast.
2. A short, professional vendor email for the issue I named (reorder, price increase, short or wrong delivery, or substitution).

Rules: base every quantity ONLY on the pars, on-hand, and forecast I gave you — do not invent usage rates or lead times. Never agree to a price change, minimum, or contract term on my behalf. Anything that commits money or terms, mark [APPROVE BEFORE SENDING]. Keep the email firm but collaborative, and short enough to read on a phone.

Fill in your details and the prompt updates live — then copy.

Common questions from restaurant managers

Is it allowed for restaurant managers to use AI like ChatGPT?

Yes — no law prohibits using AI to draft schedules, replies, SOPs, or marketing. The risk is in what you feed it and what you publish. Keep employee and guest personal data out of consumer tools, and keep the final calls on labor law, food safety, and legal matters with a human.

Can I use AI to build my staff schedule and figure out tips?

AI can draft a schedule and shift notes, but it should not decide compliance. Predictive-scheduling and fair-workweek rules vary by city and state, and the Department of Labor bars managers and supervisors from tip pools (opinion letter FLSA2024-02). Have AI draft, then verify overtime, posting deadlines, and any tip math with payroll or an employment attorney.

What employee or customer information is safe to paste into AI tools?

As little as possible. Leave out Social Security numbers, wages, home addresses, full legal names, and any guest payment or card data — consumer tools may retain what you type. First names or initials, roles, dates, and figures are enough to get a good draft.

Is it against the rules to use AI to write review responses?

Replying to reviews with AI help is fine. What is not allowed: the FTC's 2024 rule bans fake or undisclosed reviews, so never generate, buy, or solicit fake reviews, and never have staff post reviews without disclosing they work there. Any public reply must be truthful and stick to what you have actually confirmed.

Related professions