Coffee Cart Business Plan: Costs, Pitch and Menu Built With ChatGPT

Work in four passes inside one ChatGPT Project: list every start-up cost and replace the guesses with quotes, model each pitch fee per trading day, cost every cup including the card fee, then divide daily fixed costs by contribution per cup to get break-even cups. Check permits with your local licensing office, not the chatbot.

ChatGPT is good at the shape of a coffee cart business plan: it lists what you forgot, drafts formulas and writes the pitch. It is poor at the facts of your plan, because it does not know your supplier quotes, your pitch fees or your local rules, and it will fill any gap with a plausible number. The way to use it is to label every figure as a quote, an estimate or a placeholder, and to treat the plan as unfinished until no placeholders remain.

Follow me on Instagram@sagnikteaches

Set up one ChatGPT Project so the plan stays consistent

A Project keeps your instructions and files together, so you stop re-explaining the cart every time you open a new chat. Projects work on the free plan, with a smaller file allowance than paid plans get; Go costs $8 a month and Plus $20, and any of them will do for this job. Use a Project rather than a custom GPT: OpenAI has stopped new custom GPTs and existing ones stop running on 11 December 2026. Add every quote, organiser fee sheet and supplier price list to the Project's files as you collect them.

Connect on LinkedInSagnik Bhattacharya

Start with instructions that force honesty about numbers. These are the ones to paste into the Project's instruction box, with your own cart details at the bottom:

Subscribe on YouTube@sagnikteaches

You are helping me plan a one-person coffee cart. Rules for every answer:
1. Label every number QUOTE (from a document I have uploaded), ESTIMATE (my own guess,
   which I will state) or PLACEHOLDER (you suggested it). Never present a placeholder as fact.
2. Do not state any law, permit rule or fee. Tell me to check with the local licensing office.
3. Show formulas, not just results, written for Google Sheets.
4. Use USD. Margin = (price - cost) / price.
5. If two answers differ, say which assumption changed.
Cart facts: trading 3 days a week, 6 hours a day, about 40 weeks a year.
Menu: espresso, flat white, latte, iced latte, filter coffee. Card payments only.

The first rule does most of the work. Once every figure carries a label, the plan shows its own gaps, and whatever still says PLACEHOLDER is your to-do list. The second rule stops the model reciting permit rules from memory, which it will do fluently and sometimes wrongly. On Plus and above, ChatGPT Work can also build the spreadsheet file itself; on any plan you can paste its tables into your own sheet. Either way, the checks later in this tutorial apply.

Start-up costs: ChatGPT finds the gaps, quotes fill them

Ask for a checklist, not a total. If you ask ChatGPT what a coffee cart costs, it produces a confident number built from nothing. If you ask what you need, it behaves like a well-read friend who has seen plenty of carts.

List everything a one-person coffee cart needs before its first sale, grouped as
equipment, water, power, safety, paperwork and branding. For each item give a one-line
reason and a column called "quote status". Do not give prices.

Illustrative output (abridged)
Equipment: espresso machine and spare gasket kit; two grinders; knock box; jugs; thermometer
Water: fresh tank; waste tank; pump; hand-wash basin; filter and spare cartridges
Power: generator or inverter; cable covers; fuel or gas storage
Safety: fire extinguisher; first-aid kit; weights for the canopy; non-slip mat
Paperwork: insurance certificate; permit copies; training certificates; invoice folder
Branding: menu board; price display; allergen information card

Compare that with your own list. The unglamorous lines tend to go missing from a first draft: the hand-wash basin, wind weights, spare machine parts and cable covers. Then replace each line with a real quote or listing price and set its status. The table shows the structure with placeholder amounts so you can see the sums. These are illustrative figures for one possible cart, not a guide to what yours will cost.

The contingency is a planning habit rather than a rule, so pick your own percentage, but never zero. If you are buying a used machine, ask ChatGPT for a list of questions to put to the seller, such as service history, boiler age and what a replacement pump costs. It cannot inspect the machine, so the answers have to come from the seller and from your own eyes.

Pitch fees: three fee shapes and where each is cheapest

A pitch is the spot where you trade, also called a stall, booth or vendor space, and its fee is easy to misjudge. Organisers quote fees in different shapes that cannot be compared at a glance: a flat fee per day, a percentage of your takings, or a minimum plus a percentage. Ask for the full cost of a day, including any application fee, deposit, power hire and insurance requirement, and count the time you spend setting up and packing down. For the general stall version of this exercise, the street food stall business plan covers pitch fees, margins and stock.

The cheapest shape changes as takings rise, so the comparison needs your expected takings. Give ChatGPT three structures and three levels of daily takings:

Compare three pitch fee structures at daily takings of $450, $750 and $1,200:
(A) flat $85; (B) 12% of takings; (C) $40 plus 6% of takings.
Show each fee in dollars and as a percentage of takings, and the takings level at which
each pair of structures costs the same. Write the formulas for Google Sheets.

The flat fee costs the same as the 12% fee at takings of about $708 a day. The hybrid ties the flat fee at $750 and the percentage fee at about $667. So a percentage fee protects you on slow days and punishes busy ones, a flat fee does the reverse, and the hybrid sits between them. If you are not confident of reaching $700 a day, the percentage shape is the safer choice for a first season. Once you know your real level, you are in a position to negotiate a flat fee.

Whichever shape you are offered, ask each organiser the same questions and record the answers in the sheet:

• Is there a trial day, a discount for committing to a season, or a refund when weather closes the event?
• Is power or water provided, and what does it cost if so?
• Which other coffee or drink sellers trade there, and does anyone get exclusivity?
• How many people attend on a typical day, and how is that counted?
• What do you need to show: an insurance certificate, a hygiene certificate, product photos?
• When are applications due, and are deposits refundable?

When the application forms arrive, ChatGPT is a fast drafter; drafting vendor and festival applications with AI shows how to keep the facts in your hands while it does the wording.

Menu margins: cost one cup, then add the card fee

Each drink needs a recipe card listing grams of beans, millilitres of milk, and the cup, lid, sleeve and any ice or syrup. Multiply by your supplier's unit prices and you have the cost of one cup. The detailed method, with milk and syrup costs and price rounding, is in coffee shop pricing with cost per cup. The cart version adds the cost that bites hardest on small tickets: the card fee.

Square's pricing page lists in-person card payments at 2.6% plus 15 cents on its free plan. Providers differ, so use the rate you have actually been quoted. The fixed 15 cents is what matters on a cart. On a $3.00 espresso the fee is 23 cents, 7.6% of the price. To keep the plan cautious, assume one cup per card payment; bigger baskets dilute the fixed part. If your pitch has patchy signal, check how the card reader behaves offline before you buy it.

All figures are illustrative placeholders. The filter coffee earns the thinnest margin because its price is low against the fixed cost of the cup and the fee, while the iced latte does best because ice is cheap and the price steps up. Now weight the drinks by how many you expect to sell. With a mix of 10% espresso, 30% flat white, 28% latte, 17% iced latte and 15% filter, the average price is $4.36, the average ingredient and packaging cost $0.97 and the average card fee $0.26. That leaves a contribution of $3.13 per cup, the number the break-even sum needs.

Break-even cups per trading day

Break-even is the fixed costs of a trading day divided by contribution per cup. Fixed costs here means everything that does not change with the number of cups sold: the pitch fee, insurance and licences, fuel and transport, and the start-up cost recovered over three years of trading. Whether your own pay counts as a cost is a decision, and for a plan you might show to a lender, it should. The recovery period and trading days are assumptions you can change in one cell.

At $3.13 per cup, covering $172.47 takes 56 cups and covering $292.47 takes 94 cups. The recovery line includes the first year's insurance and permit costs from the start-up table, which the insurance line counts again, so year one is slightly cautious. Over a six-hour day, 94 cups is about 16 an hour. Whether that is comfortable depends on two things the sheet cannot know: how fast you can serve and how many people will come. Time yourself making twenty drinks back to back at a pace you could hold for an hour, and ask the organiser about attendance. For the general method behind the sum, break-even analysis with AI for a new product works through a second example.

Quiet, normal and busy days

The picture is sobering, and that is the point of drawing it. On a normal day this cart pays you its target wage and not much more; on a quiet one it loses money, and a plan that only works on busy days is a risky plan. The scenarios are also where weather, school holidays and winter go. Run the quiet case for the first eight weeks until real cups replace the guess.

Levers worth testing before you buy anything

Raise prices by 50 cents a cup and contribution per cup rises to about $3.62 (the card fee takes a cent of the extra), so break-even falls to 81 cups. Selling more large drinks lifts the average ticket too, but the extra milk and the bigger cup eat part of the gain, so cost the large size before you count on it. Negotiate the pitch fee from $85 to $60 and break-even falls to 86 cups. Together the price rise and the cheaper pitch take break-even from 94 cups to 74, which is far easier than finding 20 more customers a day. Ask ChatGPT to run each change separately and then together, and read the formulas it uses.

The permit step: ChatGPT drafts questions, your local office answers

This is the step where a chatbot is most tempting and least trustworthy. Rules for street trading, food businesses, water, gas, generators, hygiene training and insurance vary from place to place and change over time, and a model has no way of seeing yours. Use it to make sure you ask everything, then take the answers from the people who decide.

I plan to sell coffee from a mobile cart at markets and events. Without stating any rule,
list the questions I should ask (1) my local licensing or trading office, (2) the food
safety or health office, (3) market and event organisers, (4) my insurer.
Group them by topic and add columns for "answer", "source and date" and "fee and renewal date".

Illustrative output (abridged)
Licensing office: Do I need permission to trade from a cart at a fixed pitch, at events, or both?
  Is permission tied to a place or to me? How long does approval take and what does it cost?
Hygiene office: Must I register as a food business before trading, and how early?
  Is a training certificate needed, and at what level? Is there an inspection, and what does it cover?
Organisers: Is public liability insurance compulsory, and at what level of cover?
  Is power supplied, or may I run a generator? Are there rules on waste water and noise?
Insurer: Does the policy cover events? Machine, stock and fuel? Any conditions on towing?

Send the questions as a short email and keep the replies. A draft that works for the first message:

Subject: Questions about trading a coffee cart at markets and events
Hello,
I plan to sell espresso-based drinks and filter coffee from a small mobile cart, [two or three] days
a week at markets and events. Could you tell me (1) what permission I need to trade and whether it
covers both events and fixed pitches, (2) the fee and how long approval takes, (3) any requirements
for water, waste water, gas or a generator, and (4) whether I must register as a food business and
which training certificate is needed? A written reply would help, as I am preparing a plan.
Thank you, [your name]

Paste each answer, with the date and who gave it, into the Project and the permit tab of your spreadsheet. Anything unanswered stays marked PLACEHOLDER, and nobody should trade on a placeholder. Where the answer costs money, such as a permit fee, a certificate or a renewal, it goes back into the start-up and fixed-cost tables, which will shift your break-even.

Turn the sheet into a two-page plan and a pitch to the organiser

Once the numbers hold, ask ChatGPT to write from the sheet rather than from imagination. A prompt that works: "Using only the figures in the attached sheet, write a two-page plan with these headings: concept, menu and prices, start-up costs, pitch options, break-even, three scenarios, permits and insurance status. Where a figure is missing, write MISSING instead of estimating." The word MISSING is the useful part, because it turns every gap into something visible, and anything labelled a placeholder in the sheet should stay labelled in the text.

The same sheet feeds the pitch you send to organisers, which is shorter and more personal. This is a template with the facts left for you to fill; do not claim a hygiene rating, insurance or an inspection result you do not hold.

Hello [organiser's first name],
I'm [your name] and I'd like to apply for a regular pitch for my coffee cart. I serve
espresso-based drinks, iced coffee and filter coffee from a [length] by [width] cart. My menu
runs from $3.00 to $5.25. I bring my own water and power, [attach insurance certificate and
any hygiene certificate], and leave the pitch clean. I'd like to start with [two] trial days
and would be glad to talk through whichever fee structure suits you best.

For a lender or investor the sheet leads instead: start-up total, break-even cups, the three scenarios and the permit status. Ask ChatGPT to name the weakest assumption in the plan and to say what evidence would settle it; an answer such as "the normal-day cups figure rests on one guess about attendance" tells you where to spend your first four trading days. The prompts in a filled-in business plan prompt set for a bank go further if you need a formal document.

Trial days: a log that corrects the plan

Every number in the break-even section is a guess until the cart has traded. Keep a one-line log for each trading day, and feed it back to the Project after the first two weeks. This is what six days might look like, with the pitches reduced to two types:

Here are six trading days. Columns: day, pitch type, weather, hours, cups, takings.
[paste the rows]
For each pitch type and weather, give average cups, cups per hour and takings. Compare each with
my plan (normal day 100 cups, quiet day 60 cups, average ticket $4.36). Say which assumptions the
data supports, contradicts or cannot test yet, and how many trading days would settle each one.

Illustrative output
Market, dry (2 days): 100 cups, 16.7 an hour, takings $436.
Market, wet (1 day): 58 cups, 9.7 an hour, takings $253 (58% of a dry market day).
Event, dry (2 days): 136.5 cups, 22.8 an hour, takings $597.
Event, wet (1 day): 74 cups, 12.3 an hour, takings $322 (54% of a dry event day).
Average ticket across all days: $4.36, matching the plan. Supported.
Normal day of 100 cups: supported for markets, but only 2 dry days (thin).
Quiet day of 60 cups: close to the wet market day; one data point.
Cannot test yet: winter volumes, school holidays, new-pitch ramp-up.

Two things follow. The quiet case in the plan now has evidence behind it: wet days ran at roughly 55% of dry days, so any forecast-rain day should be planned at that level. And the pitch fee looks different with real takings beside it. The flat $85 market fee took 19.5% of takings on a typical dry day and 34% of the wet day's $253, while a 12% fee would have cost about $30 on that wet day. If the market is open to negotiation, that is your argument for a percentage or hybrid fee in the first season. Event days ran at nearly 23 cups an hour on dry days, so check you can serve that pace without a queue forming.

Where ChatGPT's arithmetic goes wrong, and how to catch it

The model writes sums fluently and sometimes wrongly, and the wrong ones tend to look plausible. Three slips are easy to make in a plan like this, and each is easy to catch once you know its fingerprint. The figures here are illustrative.

• The wrong denominator. A first answer said break-even was 79 cups, not 94. It had spread the start-up recovery and insurance over 365 days instead of 120 trading days, which makes every day look cheaper. Ask "what number did you divide by?" for any per-day figure.
• The missing card fee. Leave the fee out and contribution per cup looks like $3.39, giving a break-even of 87 cups. The fingerprint is a margin that is too close to the ingredients-only margin; check that every cost line in the sheet appears in the formula.
• The doubled line. The first start-up list totalled $19,300 against lines that added up to $18,850, because "smallwares" and "barista tools" were counted as separate items. The cure is to add the lines yourself.

Four habits catch nearly everything. Ask for formulas, then paste them into your own sheet so the sheet does the arithmetic, not the chat. Recompute three numbers by hand: a total, a per-cup cost and the break-even. Change one input, such as the pitch fee, and check that the outputs move the way common sense says they should. And ask for a list of every assumption it made and every number it supplied itself, then compare that with your quotes. If you want more practice with prompts for formulas, using ChatGPT to write Excel formulas has worked examples.

Finally, treat the plan as a living sheet. After your first trading days, replace the three scenarios with real cups per hour, re-run the break-even and see whether the lever tests still point the same way. A plan that has met real customers is worth more than one that has only met a chatbot.

Questions about planning a coffee cart with ChatGPT

Can ChatGPT tell me which permits a coffee cart needs?

It can list the kinds of permission to ask about, such as trading, food business registration, hygiene training, insurance and market rules, and it can draft the email. It cannot know your local rules or whether they have changed. Treat every answer as a question to put to your local licensing office, and keep their written reply.

How much does it cost to start a coffee cart?

It depends on the cart, whether equipment is new or used, the water and power set-up and local requirements, so no single figure is reliable. The worked example here totals about $21,000 including a 12% contingency, but every line is a placeholder. Replace each one with a quote before you rely on the total.

Is the free version of ChatGPT enough for this plan?

For drafting lists, building formulas and checking arithmetic, yes. Paid plans mostly add higher usage limits and room for more files in a Project. Whichever plan you use, do not paste bank details or identity documents, and verify every formula in your own spreadsheet instead of trusting figures typed in the chat.

How do I know my cups-per-day estimate is realistic?

Ask the market organiser for typical attendance and how many other drink sellers trade there, then treat your estimate as a guess to test. Use the first four trading days as data collection. Compare break-even with the cups per hour you can actually serve, which you can time by making twenty drinks back to back.

Should I build the plan in Excel or Google Sheets?

Use whichever you already know. Ask ChatGPT for formulas written for that program, keep every assumption on one tab, and let the other tabs refer to it. When a quote or fee changes you then edit one cell instead of hunting through the plan. Either tool handles everything in this tutorial.

Further reads

• Can AI Write a Business Plan for a Food Truck? A Worked Example — A worked business plan for a mobile food business.
• Food Truck Pricing: Set Menu Prices After Fees, Fuel and Pitch Costs — Set menu prices after fees, fuel and pitch costs.
• Pop-Up Shop Business Plan: Stock, Rent and Staffing in ChatGPT — Plan stock, rent and staffing for a short-term pop-up.
• How Can a Coffee Shop Use AI? Start With Supplier Prices — Use supplier prices to cost a coffee menu.
• Food Truck POS: Offline Mode, Fees and Which AI Features Pay Off — Offline mode, fees and which till features pay off for a mobile business.
• How to Write a Business Plan With AI, and What to Check Yourself — The general method for AI-assisted plans and what to verify.
• What Is Cash Flow Forecasting? A Plain-English Guide With AI Examples — Turn the plan into a month-by-month cash forecast.
• Board Game Café Business Plan: Table Pricing, Menu and Bookings With AI — How a board game café can price table time, cost a game-friendly menu, design booking slots and plan year one, with AI drafting and a spreadsheet checking.
• Starting a Business Using AI: A First-Month Plan for a Shop or Cafe — Four weeks before opening: names, supplier emails, menu or range, a cash-flow draft and launch posts, with the checks a person must do.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Square pricing page (in-person card rate on its free plan); OpenAI help pages on ChatGPT Projects and custom GPT retirement, as recorded in September 2026. Every cost, fee and volume figure in the examples is an illustrative placeholder.