How Food Trucks Can Use AI to Plan Stock, Pitches and Posts

Keep a one-line log for every trading day (pitch, weather, crowd size, fee, sales by item, what sold out), then once a week give it to an AI assistant to suggest prep quantities per pitch, rank pitches by profit per hour after fees, and draft your "where we are this week" posts. The log does the work; AI just reasons over it faster.

Food trucks are harder to forecast than restaurants because every day is a different location, crowd and forecast, and a busy truck might trade 150 days a year. That's thin data. An assistant can make sense of 30 rows in seconds, but it will also read patterns into three of them. The trick is to ask for ranges, tell it when you sold out, and treat its numbers as a starting point for your own judgement.

Follow me on Instagram@sagnikteaches

The trading log: the columns that make everything else work

Keep it in a spreadsheet on your phone and fill it in while packing down, when the day is fresh. Nine columns cover it:

Connect on LinkedInSagnik Bhattacharya

Here's what a few filled-in rows look like for an illustrative burger and loaded-fries truck:

Subscribe on YouTube@codingliquids

Sat 6  | Riverside market | weekly market | warm, sunny | ~2,000 (org.) | $60 flat
       | 212 covers | burgers 96, fries 150, veg burger 31 | fries sold out 14:40 | 3 buns
Tue 9  | Business park    | lunch         | cool, dry   | ~400 (guess)  | $25 flat
       | 71 covers  | burgers 38, fries 55, veg burger 14  | none            | 6 buns, 2kg chips
Sun 14 | Summer festival  | festival      | hot, sunny  | ~9,000 (org.) | $450 + 10%
       | 540 covers | burgers 260, fries 410, veg burger 88 | burgers 19:10  | none

Your till does most of the work here: nearly every card-payment system can export sales by item for a date range. How AI inventory forecasting works for small businesses explains what the forecasting side does with data like this.

Stock: prep quantities for each pitch

Once you have six to eight weeks of rows, ask for a prep plan for the coming week. The prompt matters, because the default answer is a single confident number:

Attached is my food truck's trading log. Next week I'm trading at:
[list pitches, dates and the weather forecast for each].

For each day, suggest how many burgers, fries portions and veg burgers
to prep, as a low / likely / high range.
Important: on days marked "sold out", real demand was HIGHER than the
sales figure, so don't treat those sales as the ceiling.
Tell me how many past days each suggestion is based on, and say when
the evidence is too thin to be useful.

An illustrative reply:

Sat, Riverside market (8 comparable Saturdays, good evidence)
  Burgers 95 / 108 / 120   Fries 150 / 175 / 195   Veg 28 / 33 / 38
  Fries sold out on 3 of 8 Saturdays, so I've set fries above past sales.

Tue, Business park (6 Tuesdays)
  Burgers 32 / 38 / 44     Fries 48 / 55 / 62      Veg 11 / 14 / 17

Sun, Harvest fair (no past data at this event)
  Based on the summer festival scaled to the stated crowd of 3,000:
  Burgers 85 / 95 / 105 ...

The Saturday and Tuesday numbers are reasonable. The Sunday one needs fixing. The assistant has scaled a hot-weather summer festival by crowd size to a fair it knows nothing about, and presented the result with the same confidence as the Saturday market. Treat any pitch with no history as a judgement call: prep to the low end of what you think, carry extra frozen stock if the van allows, and log the day carefully so next year you have a row to work from.

Turning the ranges into a shopping list is a spreadsheet job, and one rule keeps waste down: buy perishables to the "likely" figure and hold frozen or long-life stock to the "high" one. For the Saturday market, with illustrative quantities, that works out as:

If the high figure for frozen stock won't fit in the van, that is useful to know too: it tells you which item will sell out first, and roughly when, before you've left the yard.

Plans made on Sunday also need a second look when the forecast turns. Say Friday's forecast for the market changes to heavy showers until 1pm. Re-run just that day: "Re-run Saturday's Riverside market for heavy showers until 1pm, then dry. How many past market days had rain, and how did sales compare?" An illustrative reply: burgers 70 / 80 / 92, based on two wet Saturdays, "which is thin evidence; both wet days lost most of their trade before noon". Two rows is not a forecast, but it is enough to cut the bun order by two packs and keep the frozen stock as the buffer.

The sold-out instruction is the detail most people miss. If fries sold out at 2.40pm on a market that runs until 4pm, you didn't sell 150 portions because 150 people wanted them; you sold 150 because that's all you had. A rough sum gets you close to the real figure. The market opened at 10am, so 150 portions went in 4 hours 40 minutes, about 32 an hour. If the last 80 minutes usually run at around two-thirds of the morning pace, that's another 28 or so portions the truck could have sold, putting real demand near 178 rather than 150. The assistant's "likely" figure of 175 is in the same place, which is a good sign it took the sold-out note seriously. Bakeries face the same problem with unsold and sold-out lines, and how bakeries use AI to predict demand shows the same correction applied to a counter.

Pitches: which spots and events are worth the fee

Takings flatter big events. The fairer comparison is profit per hour of your time, counting setup and pack-down, after food cost, fees, fuel and any extra staff. Here are the three pitches from the log, with illustrative figures and a 32% food and packaging cost:

The festival takes nearly three times as much as the market but earns less per hour, and carries far more weather risk. The business park is the weakest, but it fills a Tuesday that would otherwise earn nothing, so it may still be worth keeping. The arithmetic belongs in your spreadsheet; ask AI afterwards to spot what the table doesn't show, such as which pitches only work in good weather or which have a falling trend.

Check its trends against what you remember. In an illustrative run, the assistant flagged the business park as "declining: 71, then 58, then 49 covers over the last three Tuesdays" and suggested dropping it. The log's weather column told a different story: the two later Tuesdays were cold and wet, and the one dry Tuesday in between had matched the first. The trend was the weather. The prompt that avoids this asks the assistant to compare like with like: "Compare only days with similar weather before calling anything a trend."

"Pitch" has a second meaning for trucks: the application you send to an event organiser. AI is handy here. A before and after for an illustrative application paragraph:

Before: We are a burger truck and would love to come to your event.
We have lots of experience and great food.

After: We serve smash burgers and loaded fries from our own truck,
at about 60 covers an hour at peak with two staff. Last season we
traded at eight weekend markets and two festivals of 5,000+ visitors.
We carry our own power, bring a veg and a gluten-free option, and
take card payments only. Photos of the truck and menu are attached.

Organisers pick traders who make their event easier to run. Throughput, power, dietary options and payment method answer the questions they'd otherwise have to ask. Give the assistant your real numbers; don't let it invent a track record.

Many organisers use an application form rather than an email, and the same facts fill it. Keep a trader sheet with the answers written once, so each form takes minutes. An illustrative, filled-in extract:

Trading name and cuisine: burger truck; smash burgers, loaded fries
Unit size: 6 m long x 2.5 m wide, serving hatch on the left side
Power: self-sufficient (quiet inverter generator); can use a 16A
  hook-up if provided
Water and waste: own fresh and grey water tanks; we take all
  waste and cooking oil away
Throughput: about 60 covers an hour at peak, two staff
Dietary options: veg burger; gluten-free bun on request
Payments: card and phone only
Documents attached: liability insurance, food business
  registration and latest inspection report, gas safety and electrical test certificates, menu with prices

When an organiser asks something new, add the answer to the sheet. Ask the assistant to tailor the covering paragraph to each event; leave the sheet itself alone, because those facts don't change from one application to the next.

Posts: telling people where you'll be

For a truck, the most valuable post is the boring one: where and when. Three posts a week cover most of it:

• The week-ahead schedule, posted on Sunday or Monday, listing every pitch, day and time.
• The day-of post, with a photo of today's special and the exact spot within the site ("by the south gate, next to the coffee cart").
• The sold-out or last-hour post, which rewards followers who come early and quietly advertises demand.

Meta Business Suite is free and lets you schedule posts to Facebook and Instagram in advance, so the week-ahead post can be written and queued on Sunday night. An illustrative AI draft of that post, and what the owners fixed:

This week's stops:
Tue 16: Business park, 11.30am-2pm
Sat 20: Riverside market, 10am-4pm (by the south gate)
Sun 21: Harvest fair, 11am-5pm
Every Tuesday we're at the business park, so pop by for lunch!
#foodtruck #burgers #streetfood #smashburger #loadedfries
#foodie #lunch #weekend #market #eatlocal #yum #foodlover

Two fixes. "Every Tuesday" isn't true, since the business park now runs on fair-weather Tuesdays only, and a follower who turns up in the rain will remember it. And Instagram has capped hashtags at five per post since December 2025, so the twelve became five: the three that describe the food, plus the market and fair names people actually search. Ask for "no more than five hashtags" in the prompt and you won't have to cut them.

A before and after from the day-of post shows what AI does well:

Before (typed in a hurry): at the market today come find us

After (AI draft from the log and your notes): Riverside market today,
10am-4pm, by the south gate. Special: the chilli-honey smash burger.
Fries went by 2.40 last Saturday, so come early if they're your thing.

A realistic mistake: an assistant asked to "write this week's schedule post" from a list pasted in a hurry wrote "see you at the market from 11am", because an earlier message in the same chat mentioned an 11am start at a different pitch. Two dozen people arrived an hour after the truck opened and a few left comments. Always check day, date, time and place against the log before scheduling. A pre-publish checklist for AI-drafted posts covers the other things to look at, and writing Instagram captions that sound like you helps if the drafts start sounding generic.

A Sunday-evening planning routine

All of this fits into about 45 minutes once a week:

1. Five minutes: check the week's log rows are complete, especially sold-out times and waste.
2. Ten minutes: paste the log and next week's pitches and forecast into the prep prompt. Adjust the ranges with your own judgement and turn them into a shopping list.
3. Ten minutes: once a month, update the pitch profit table and ask what's changed.
4. Fifteen minutes: draft and schedule the week-ahead post, and draft the day-of posts to finish on the day with a fresh photo.
5. Five minutes: note anything odd about next week (a road closure, a clash with a bigger event) in the log so it's there when you look back.

The cost is small. A spreadsheet and your till's reports are free, Meta Business Suite is free, and a general assistant is about $20 a month for ChatGPT Plus or Claude Pro, though a free tier may cover a weekly routine this size. The saving shows up in two places: less waste on quiet days, and fewer sold-out afternoons on busy ones.

The same routine looks different for a coffee van. Weather cuts the other way (hot afternoons sell iced drinks and fewer flat whites), the busiest hour is often before the market officially opens, and milk is the stock that runs out. The log columns stay the same, but "sales by item" should split hot and cold drinks, and the prep prompt should ask about milk and cups rather than buns. Tell the assistant what kind of truck you run in the first line of every prompt, or it will reason like a burger van.

What changes in a month for a two-person truck

Say the burger truck above trades 14 days in a month: eight weekend markets and festivals, six weekday lunches. Before keeping the log, the owners prepped from memory. Quiet weekdays ended with about $55 of food thrown away or taken home, and weekend days sold out of something before closing on five of eight days.

After a month on the routine, weekday waste averages about $30, a saving of roughly $150 across six weekdays. Early sell-outs drop to two of eight weekend days. If each of the three avoided sell-outs was worth around 15 extra covers at $12.50 each, that's about $560 in sales they'd previously turned away. The pitch table also shows that one weekday lunch spot is earning less than the fuel and fee justify in bad weather, so they switch it to fair-weather days only.

None of that needed special software. It needed a log filled in every day, sold-out times recorded honestly, and 45 minutes on a Sunday. The assistant made the weekly analysis quick enough that it actually happened.

How a weekly plan can mislead you

• Trusting a forecast built on three rows. Ask how many past days each number rests on, every time.
• Feeding it sold-out sales as if they were demand. Mark sold-out times, and tell the assistant what they mean.
• Menu changes the log doesn't mention. Add a chicken burger in week five and it takes sales from the beef burger, but an assistant reading the log keeps forecasting beef at the old level. Write "menu change: chicken burger added" in the notes on the first day, and tell the assistant to treat earlier weeks as a different menu.
• Organiser crowd figures taken at face value. "9,000 visitors" is sometimes a whole weekend's tickets, not one day's crowd. Ask the organiser which it is, and log the answer, or every forecast scaled from it will be too high.
• Stale forecasts. A plan made on Sunday with a sunny forecast doesn't survive a wet Friday. Re-run the day you're worried about the night before.
• Posts with the wrong time or place, usually from a long chat that mixes several weeks. Start a fresh chat each week and check every post against the log.
• Letting AI decide which events to drop. A pitch that earns less per hour might bring you private-hire bookings or keep a good organiser relationship. The table informs the decision; you make it.

If you're unsure how much of your social media to hand to AI at all, whether a small business should use AI to write its social posts sets out where it helps and where your own voice matters more.

Further reads

• How Small Shops Can Use AI to Forecast Stock and Reorder — Reorder planning for the ingredients your prep plan depends on.
• How to Use AI to Set Reorder Points and Prevent Stockouts — Setting reorder points so the van never runs short on busy weeks.
• Best AI Social Media Tools for Small Businesses (2026) — Scheduling tools for the weekly schedule and day-of posts.
• Can AI Run Your Social Media on Autopilot? What Breaks — What breaks if you let AI post without checking.
• How to Write Menu Descriptions With AI That Still Sound Like You — Menu board and listing wording that sounds like you.
• Can AI Help a Small Shop Set Prices and Promotions? — Pricing specials and bundles for slower pitches.
• How to Forecast Covers and Cut Food Waste With AI — A step-by-step covers forecast a small restaurant can build in a spreadsheet with AI help, turned into prep quantities and checked against a daily waste log.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: ChatGPT and Claude plan pricing pages; Meta Business Suite help on scheduling posts, checked September 2026.