You know the feeling. You wrapped a job last week that came out clean, the customer was thrilled, the check cleared without a hitch. And a small voice keeps asking whether you actually made a dime on it.
Most of us quote by gut. You eyeball the work, you remember what the last one like it ran, you add a little cushion, you send the number. Then real life shows up: the material bill runs higher than you figured, a guy calls out, you drive back twice for a punch list nobody's paying for.
By the time the job's closed, the profit is a fog. Your bank balance looks fine because three jobs are blurred together, and the one that quietly lost money is hiding inside the two that did well. Here's a way to pull a single job apart in about ten minutes with an AI chatbot, and start seeing which work is actually worth loading the truck for.
Start with the four numbers that decide a job
Every job's profit comes down to a handful of inputs, and most of them never make it onto paper. To get a margin you can trust, you need four things in front of you.
- Labor hours, all of them. Not just the billable ones. The hour spent loading the truck and the forty minutes at the supply house count too.
- Materials at what you actually paid. Tax included, not the list price you meant to charge.
- Drive time, both directions. Plus fuel. A job two towns over eats more than one down the street.
- Callbacks and rework. The return trips nobody billed for. This is the one that hides.
Notice what's missing from most people's math: your own time. It isn't free just because you don't cut yourself a paycheck for it. Put a rate on your hours, even a rough one, or every job will look better on the screen than it did in your wallet.
A ten-minute costing session
There's no special software needed for this. Any AI chatbot (ChatGPT, Claude, Gemini, the free versions handle it fine) can do the arithmetic once you hand it the pieces. The flow goes like this.
- Snap a photo of your material receipts and upload it. Ask the chatbot to read the line items and total them with tax. Most tools pull numbers off a clear receipt photo well enough for a rough count.
- Type in your rough hours and rates. Something like: two guys, six hours each at $35 an hour, plus four of my own hours at $60.
- Add drive time and fuel. Forty minutes each way, two trips, and figure a fair per-mile cost for the round trip.
- Give it the invoice total, the number you actually billed the customer.
- Ask the plain question: what's my margin on this job, in dollars and as a percentage, and show your work.
Ask for the math so you can catch a bad number. If the material total looks off, the receipt photo was probably blurry, so type that figure in by hand and keep going. The whole thing runs about ten minutes the first time and less after that.
Turn it into a habit
Once you've run it a single time, copy the prompt somewhere you'll find it, a notes app or a saved message to yourself. Next job, you paste it, swap the numbers, and you're done before the coffee's cold.
Reading the answer, and spotting the quiet losers
One job's margin is interesting on its own. The pattern across ten jobs is where the real money hides, so cost a handful over a couple of weeks and then look at them together.
After you've got five or six, paste the totals back in and ask the chatbot to sort them. You're hunting for three patterns.
- By job type. Maybe your core service clears a healthy margin while the little quick-favor jobs barely break even once drive time is in.
- By customer. The friendly folks who always tack on one more thing while you're there can be your least profitable stop, and you'd never feel it without the numbers.
- By season. A January service call with a nor'easter in the forecast costs you more in time and hassle than the same call in June. Price accordingly.
What you're after isn't a perfect set of books. It's the one pattern that changes your next quote, like nudging small jobs up by fifty bucks or adding a real trip charge past a certain distance.
Where this stops (worth being honest)
This is a back-of-napkin estimate, and it's only as good as the numbers you feed it. Lowball your hours and the job looks great on the screen and lousy in the bank, same as always. A few things it flat out won't do:
- It isn't bookkeeping. It doesn't track your accounts, reconcile anything, or remember last quarter.
- It won't handle taxes. Sales tax, deductions, quarterly filings, that's your accountant's job, and this doesn't replace them at tax time.
- It ignores overhead unless you say so. Insurance, the phone bill, the truck payment, all of it gets spread across your jobs, and a raw margin skips it.
- It can misread a receipt. Faded ink and crumpled paper trip it up, so always eyeball the total it gives you.
Used for what it is, a fast gut check between jobs, it earns its ten minutes. Treat the output as a nudge for your next quote, not a ledger you'd hand the IRS.
Worth doing this week
Nobody's asking you to overhaul how you run things. Pick one finished job you remember well, ideally a recent one, and start there.
- Cost that one job tonight. Receipts, hours, drive time, invoice total, ten minutes.
- Set your own hourly rate on paper first, so your time actually counts.
- Run two more jobs of different types, so you've got something to compare.
- Save the prompt as a template you can reuse without retyping.
- On your next quote, adjust the number based on what you saw. That's the whole point.
If you'd rather have this wired into your invoicing so the margin lands in your inbox automatically, that's the kind of setup we help small shops put together. Either way, the first honest number you run will probably surprise you.