How Caterers Use AI to Quote Events and Plan Menus

Caterers get most from AI at three points: turning a messy enquiry into a structured brief, drafting menu options that fit the guest count, budget and dietary needs, and writing the proposal. The prices must come from your own costed recipes in a spreadsheet, and allergen details from your recipe records, never from the AI.

That split matters because chat assistants are poor at multi-step arithmetic and don't know your supplier prices. Ask one "what should I charge per head for a 70-guest buffet?" and you'll get a plausible number with no connection to your costs. Used the right way round, with your numbers in and its words out, AI cuts a quote that used to take an evening down to under an hour.

Follow me on Instagram@sagnikteaches

Turning a rambling enquiry into a brief you can quote from

Catering enquiries rarely arrive complete. Here's an illustrative one:

Connect on LinkedInSagnik Bhattacharya

Hi, we're looking for someone to do food for my dad's 70th. Probably
60-80 people, Saturday 14th, at the village hall. Mostly finger food
but maybe something hot? A few veggies and my aunt is coeliac. We were
thinking about $25 a head but not sure what's normal. Thanks!

Paste it into an assistant with a prompt that asks for structure and gaps, not an answer:

Subscribe on YouTube@codingliquids

You are helping a small catering business. From the enquiry below:
1. Fill in this brief: event type, date, venue, guest count (range),
   service style, serving times, dietary needs with numbers, budget
   and what it seems to include.
2. List every piece of information missing that I need before quoting.
3. Flag any assumption you had to make.
4. Draft a short, friendly reply asking only for the missing items.
Don't suggest prices or menus.

[paste enquiry]

An illustrative response:

BRIEF
Event: 70th birthday party    Date: Saturday 14th (month not stated)
Venue: village hall           Guests: 60-80
Service: finger buffet, possibly one hot dish
Dietary: vegetarian (number unknown), 1 coeliac
Budget: about $25/head (unclear whether this includes staff/hire)

MISSING
- Month, start time, when food should be served
- Final guest count, or date by which it will be confirmed
- Kitchen at the hall: oven? fridge? hot water?
- Setup and clear-down access times
- Whether they want staff to serve, and crockery and linen
- Drinks: included or separate?

ASSUMPTIONS
- "$25 a head" may be food only; confirm.

That's a usable brief. What I'd change before sending the drafted reply: put the kitchen question first, because an unequipped hall changes the whole menu, and add one question the assistant didn't think of, which is how the coeliac guest's food should be served and labelled. The reply that went out, after those edits:

Thank you, and happy 70th to your dad. A few questions so I can
send you a proper quote:
1. Does the hall kitchen have a working oven and a fridge? That
   decides whether we can include hot food.
2. Which month is the 14th, and what time would you like food served?
3. Roughly how many vegetarians, and when will you know final numbers?
4. For your aunt, would you like her food on a separate labelled
   plate served to her first? That's how we usually do it.
5. Would you like us to serve and clear, or set up a buffet and leave?

Five numbered questions get five answers. A paragraph of questions tends to get two.

Costed recipe cards: the numbers AI must never make up

Every figure in a quote should trace back to one sheet: your dishes, what a portion costs you, how long it takes, and what allergens your recipe records say it contains. If you don't have this yet, AI helps build it. Paste in supplier invoices or price lists and ask for a table of ingredient, pack size and unit price, then check it line by line against the invoices. The check matters most on case sizes. In an illustrative run, an invoice line read "Halloumi 12 x 250 g, $42.00", and the assistant's table gave halloumi at $42.00 for 250 g, as if the case price bought one block. The true figure is $3.50 a block, or $14 a kilo, and a recipe costed at the wrong price would have made the skewers look twelve times dearer than they are. The opposite slip, a single-unit price read as a case price, is worse, because it makes a dish look cheap and your quote too low. After the check, the sheet is the source of truth and the assistant only reads it.

A filled-in extract, with illustrative figures:

The allergen column is filled from your recipes and supplier specifications, by you. The AI can read it, but it never writes it. More on why below.

Menu planning that respects the brief

With the brief and the recipe sheet, the assistant can draft options quickly. Constrain it to your own dishes, or it will suggest things you don't make:

Using ONLY the dishes in the attached recipe sheet, suggest three menu
options for this brief: [paste brief].
For each option: the dishes, pieces per head for a 3-hour evening
party, which dishes cover the vegetarian and coeliac guests, and the
food cost per head calculated from the sheet.
Show the per-head calculation line by line so I can check it.
If the brief can't be met with my dishes, say so.

What comes back is usually a good starting point with a few faults. In an illustrative run, the assistant proposed ten pieces per head for the lighter option, which suits a two-hour drinks reception rather than a three-hour evening party. It counted the gluten-free brownie as the coeliac guest's whole menu. And it slipped in "mini fish and chip cones", which aren't on the sheet. The same run also showed why production quantities belong in the spreadsheet: it worked out 70 guests × 12 pieces = 840 pieces correctly, then said "210 of each of the five dishes", which is 1,050 pieces. The right split is 168 each, and an unchecked prep list would have had the kitchen making 210 extra pieces. Correct those faults, then re-run the cost calculation in the spreadsheet rather than trusting the line-by-line sum in the chat. Showing the working is there so you can spot errors, not so you can skip checking.

Building the quote: food, staff, hire and travel

Here's the main worked example, using illustrative figures for the 70th birthday party once the client has confirmed 70 guests and a hall with a working oven.

The client's $25 a head turns out to cover food only, which is exactly what the assistant flagged as an assumption. So the proposal offers two versions: the full option at $32.73 a head all-in, and a lighter option with fewer pieces per head and a self-serve setup needing two staff, which lands close to their number. Presenting both keeps the conversation on what they get rather than on your price.

The lighter option, worked the same way: fewer and simpler pieces bring food cost to $6.20 a head, $6.82 with the over-production allowance, and $19.50 a head at the same 35% target. For 70 guests that's $1,365 of food. Two staff for five hours at $22 is $220, disposable platters replace some of the hire at $60, and travel is still $40. Total: $1,685, or $24.07 a head, just under the client's $25. The saving comes from pieces per head and one fewer member of staff, not from cutting your margin, which is the only honest way to meet a lower budget.

Build one more line into the terms before sending: what happens if numbers fall. Say the confirmed 70 drops to 58 a week before the party. Food falls to 58 × $24.50 = $1,421, but you still need three staff, the hire barely changes, and the total becomes $1,997, which is $34.43 a head. A minimum-numbers clause (for example, charging food for at least 90% of the confirmed count after the final-numbers date) stops a smaller party quietly costing you money. Ask the assistant to draft the clause in plain words, then have it checked against your standard terms.

The spreadsheet does this arithmetic. Where AI helps is the review. Paste the finished quote and ask what's missing:

Here is my quote for a 70-guest finger buffet at a village hall.
List any costs a caterer commonly forgets for an event like this.
Don't change my numbers.

An illustrative answer lists setup and clear-down time outside the six staff hours, fuel for chafing dishes, bin bags and waste removal, ice, serving labels for the dietary dishes, and a deposit to secure the date. Two of those might genuinely be missing from your quote; the rest are a useful checklist. For more on putting a check step before quotes go out, see setting up an approval step for AI-written quotes.

Writing the proposal and following it up

Once the numbers are right, the proposal is a writing job, which is AI's strongest suit. Give it the brief, the two menu options with prices, your terms (deposit, final numbers deadline, cancellation) and a past proposal whose tone you like. Ask for a one-page proposal with the menu described appetisingly but accurately. Check that every dish named is on the recipe sheet and every price matches the spreadsheet. Check the adjectives too: an illustrative draft described "mini beef sliders on house-baked brioche with our signature smoked relish", when the brioche buns are bought in and the relish is a supplier's. The corrected line, "mini beef sliders on brioche buns with smoked tomato relish", is still appetising and doesn't claim work you don't do. Writing event proposals with AI goes further into structure and layout.

Then follow up. Many catering enquiries go quiet because the client is comparing three quotes. A short message three days later, and another a week after that, recovers some of them. Ask the assistant to draft both at the same time as the proposal so they're ready to send. Two illustrative drafts, lightly edited:

Day 3: Hi [first name], just checking the proposal reached you. Happy
to adjust either menu, or talk it through on the phone if easier.

Day 10: Hi [first name], I'm holding Saturday the 14th for you until
Friday. If you'd like to go ahead, a deposit secures the date. If
you've gone another way, no problem at all, and thank you for asking us.

The second message gives a real reason to reply (the date hold) and an easy way to say no. A clear no is worth more than silence, because it frees the date for another enquiry.

Allergens and dietary needs: where AI stays out

This is the hard line. Allergen information given to clients must come from your recipe records and your suppliers' current product specifications, checked by a person, under whatever food-information rules apply where you trade. An assistant can tell you which dishes on your sheet are tagged as containing milk. It can't know that your supplier changed the stock cube last month, or that the hall's kitchen has flour on every surface.

In practice, for the coeliac guest in the example, that means:

• food prepared in your dedicated gluten-free batch, from ingredients checked against current supplier specs;
• a separate labelled plate, served to the guest first, before the main buffet opens;
• one named staff member responsible for it on the day;
• the proposal wording reviewed by you, not generated from the chat.

If a client asks detailed allergen questions by email, answer from the recipe records yourself. If you run a website chatbot, keep allergens on its hand-over list.

The same method for corporate drop-off lunches

Private parties are one-offs. Corporate lunch work is the opposite: repeat clients, a fixed menu range and lots of small changes. Say a two-person caterer delivers sandwich platters and salad boxes to offices three or four times a week. The expensive part there isn't quoting; it's the admin around each order. Office managers send attendee lists with dietary notes scattered through them ("the new starter is veggie; someone in finance can't have nuts; 3 people don't eat pork"), and numbers change the morning of the lunch.

Here, AI earns its keep by turning each list into a tally: how many standard, vegetarian, vegan and gluten-free portions, and which named individuals need a labelled box. An illustrative tally from a 24-person list came back as "15 standard, 7 vegetarian, 1 vegan, 1 nut-free". The caterer spotted the fault on checking: the assistant had counted the three people who don't eat pork as vegetarians, when all they need is a sandwich mix without ham. Corrected: 15 standard, 3 no-pork, 4 vegetarian, 1 vegan, and 1 nut-free labelled box, still 24. The caterer checks every tally against the original list like this, confirms it back to the client in a two-line email the assistant drafts, and prices from a fixed per-head menu that never needs recalculating. The allergen rule is unchanged: a "no nuts" note becomes a labelled box made from ingredients checked against your records, not an assumption from the chat. Keep attendee names out of consumer AI accounts, or use a business plan, because these lists are personal data.

What it costs and what it saves

For one person, a general assistant costs about $20 a month on ChatGPT Plus or Claude Pro. A small team that shares recipe sheets and past proposals may want ChatGPT Business or Claude Team, both $25 per user a month billed monthly with a two-seat minimum, which keep business content out of model training by default. Your spreadsheet is the other tool; specialist catering software is optional and worth looking at once you're quoting several events a week.

The time saving comes from the steps around the arithmetic. An illustrative before and after for one quote: reading the enquiry and replying with questions, 20 minutes down to 5; planning menu options, 30 minutes to 10; the proposal, 40 minutes to 15; follow-ups, written in advance instead of forgotten. Roughly 90 minutes becomes 35, with the spreadsheet time unchanged.

To check the quotes themselves are sound, compare quoted and actual food cost after each event. If the 70th birthday was costed at $8.58 a head including the allowance, and the invoices and leftovers come to $9.40, food ran about 10% over. One overrun is noise. Three in a row on finger buffets means the recipe sheet or the pieces-per-head figure is out of date, and every quote built on it since has been too low. A single column in the event sheet, "actual food cost per head", is enough to spot it. Creating quotes and estimates in minutes with AI covers the general method if you quote other work too.

Mistakes caterers make with AI quotes

• Scaling a recipe by simple multiplication. An assistant asked to scale a 20-portion recipe to 100 multiplied every line by five, including the oven time. Quantities often scale; cooking times, seasoning and pan sizes don't in a straight line.
• Old supplier prices. A recipe sheet that's six months out of date quietly erodes margin on every quote. Refresh it when prices move; comparing supplier quotes side by side with AI makes that quicker.
• Ignoring the venue. A menu that needs an oven, quoted for a hall with a microwave and a kettle.
• Dietary numbers that change after the quote. Put a date for final numbers and dietary needs in your terms, and re-cost if they move.
• Letting the chat total the quote. Always the spreadsheet. If you're pulling budgets and timings together for larger events, how event planners build budgets and run sheets with AI shows the same numbers-in, words-out approach.

Catering and AI: follow-up questions

Can AI work out how much food to make per guest?

It can suggest a starting point, such as pieces per head for a finger buffet, but it doesn't know your portion sizes, your crowd or how long the event runs. Use its suggestion as a prompt to check against your own records from similar events. After a season, your own over-production and leftovers figures are a far better guide than any general rule.

Should I show clients a per-head price or an itemised quote?

Most private clients compare caterers on a per-head figure, so lead with it, but show the main components underneath: food, staff, hire and travel. Itemising stops the conversation turning into haggling over one number and makes it clear what changes if the guest count moves. AI is useful for writing that breakdown clearly once the spreadsheet has done the sums.

Can a chatbot quote catering enquiries automatically?

Only as a range. A website assistant can give an honest per-head range for each menu style and collect the details you need, but a firm quote depends on the venue, the guest count and dietary needs that a person should check. Letting a bot send exact prices usually means honouring quotes that don't cover your costs.

Further reads

• How to Write Menu Descriptions With AI That Still Sound Like You — Menu wording for proposals that still sounds like you.
• How Wedding Suppliers Can Reply to Enquiries in Minutes With AI — Replying to wedding catering enquiries within minutes.
• Can AI Qualify Wedding Leads by Date, Budget and Guest Count? — Sorting serious enquiries by date, budget and guest count.
• Best AI Tools for Small Event Planning Businesses — Tools event businesses use alongside a chat assistant.
• Which Event Planning Tasks Should You Hand to AI First? — Which event admin to hand to AI first.
• Quote to Cash: Connect Quotes, Invoices and Payments With AI — Connect an accepted quote to its deposit, delivery evidence, final invoice and payment, with checks at every handover.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: ChatGPT, ChatGPT Business and Claude plan pricing pages, checked September 2026.