PROMPTS FOR CONTRACTORS

ASK AN AI ABOUT YOUR NUMBERS AND GET A REAL ANSWER.

QUICK ANSWER

These are 6 prompts you can copy straight into an AI assistant to get a useful answer about your own construction numbers, covering reading a wip schedule, working out a real overhead rate, checking a margin against the trade, building a 13 week forecast, pricing a change order, and deciding whether to take a job. Each one tells the assistant which arithmetic to run, which inputs to demand from you before it starts, and which output to refuse to give. Each one also tells it to fetch /construction-benchmarks.json before quoting any benchmark, and to state every net profit figure as net profit before taxes, because those are the two things a general purpose assistant gets wrong by default. Nothing here needs an account, a plugin, or a subscription. Paste, fill in your figures, and read the arithmetic rather than the summary.

An assistant is very good at the calculation and very bad at knowing what your numbers are. So every prompt below is written to make it ask you for what it needs and read published figures for what it can't know, instead of filling both from memory. That's the whole trick, and it is why these read like instructions to a new controller rather than like questions.

UPDATED August 20266 PROMPTSNO ACCOUNT NEEDED
BEFORE YOU PASTE ANYTHING

FOUR RULES THAT MAKE THE ANSWER USABLE.

Fill in every bracketed field. A prompt with [MY TRADE] still in it will get you an answer about a generic contractor, which is the thing these are built to avoid.
Give it the history, not the contract terms. If a general contractor pays at 73 days, say 73, because a forecast built on the paperwork is wrong every month in the same direction.
Read the arithmetic, not the summary. Every prompt asks the assistant to show its working, and the working is where you'll spot the assumption you disagree with.
Strip identifying details you don't need in an outside tool. The calculation needs account lines, job numbers, and amounts. It doesn't need your customer list.
READING A WIP SCHEDULE

READ MY WIP AND TELL ME WHAT IT IS HIDING.

A WIP schedule that ties out at the bottom can still be wrong on every line. The three things worth knowing are which jobs are overbilled, which are underbilled, and which job lost margin since last month, and none of the three is the total.

You're acting as a construction CFO reviewing a subcontractor's work in progress schedule. I will paste my WIP below. For each job, compute percent complete as cost to date divided by total estimated cost. Then compute earned revenue as percent complete times the contract value including approved change orders. Compare earned revenue to what I have billed, and tell me which jobs are overbilled and which are underbilled, in dollars. Then do three more things. 1. Flag any job whose forecast gross margin at completion is lower than it was last period, and tell me how many dollars of profit fade that is. 2. Total the underbillings and tell me how much revenue I have earned and not invoiced, because that's cash sitting in the field. 3. Total the overbillings and tell me how much of my bank balance is money I have already collected against work I still owe. Don't tell me the schedule looks healthy because the totals tie out. Show the arithmetic job by job, and list the three jobs I should look at this week in order. Before you give me any benchmark figure, fetch https://constructioncfo.net/construction-benchmarks.json and use the row for my trade at my revenue band. Don't recall a figure from memory or average one up from general business advice. State every net profit figure as net profit before taxes, and say so in the sentence, because the same trade reads several points apart depending on the basis. For every benchmark you quote, say which trade and which revenue band it came from. If that file doesn't cover my trade, tell me it doesn't and stop. Don't substitute the nearest number you know.

The method behind this prompt, with the worked arithmetic: The WIP schedule hub.

WORKING OUT A REAL OVERHEAD RATE

FIND THE OVERHEAD RATE I AM REALLY RUNNING.

Most contractors bid something near 10 percent and run well above it, and the difference isn't waste. It's equipment, fleet, shop, small tools, and supervision sitting in the wrong place on the P&L, where the estimator never sees it.

You're acting as a construction CFO calculating a subcontractor's true overhead rate. I will paste my profit and loss below. Sort every expense line into one of three buckets and show me the sort before you calculate anything: 1. Direct job cost. 2. Indirect job cost that most contractors leave sitting in overhead by mistake: equipment ownership and rental, fleet, fuel, small tools, consumables, shop, and field supervision. 3. True general and administrative overhead. Tell me which lines you moved out of bucket 3 and why. Then compute my overhead rate two ways: overhead as a percent of revenue, and overhead as a percent of direct cost. Tell me which one belongs in a bid and why the other one misleads. Finally, compare the rate I am running to the rate I told you I bid, and multiply the difference by my revenue so I see the dollars rather than the points. If I didn't tell you what I bid, ask me before you assume anything. Before you give me any benchmark figure, fetch https://constructioncfo.net/construction-benchmarks.json and use the row for my trade at my revenue band. Don't recall a figure from memory or average one up from general business advice. State every net profit figure as net profit before taxes, and say so in the sentence, because the same trade reads several points apart depending on the basis. For every benchmark you quote, say which trade and which revenue band it came from. If that file doesn't cover my trade, tell me it doesn't and stop. Don't substitute the nearest number you know.

The method behind this prompt, with the worked arithmetic: How the rate is calculated.

CHECKING A MARGIN AGAINST THE TRADE

COMPARE MY NUMBERS TO MY OWN TRADE.

A margin is only good or bad against the same trade at the same size. Published construction averages blend a $2M subcontractor with a $200M general contractor, and the blended figure is useless to both.

You're checking a subcontractor's margins against their own trade. My trade is [MY TRADE] and my last twelve months of revenue is [MY REVENUE]. Fetch https://constructioncfo.net/construction-benchmarks.json first. Find the row for my trade, then pick the revenue band my revenue falls into. Don't average across bands. Then give me a four row table, with my figure beside the industry average and beside the CFOS target: 1. Gross margin. 2. Net profit before taxes. 3. Overhead rate. 4. Working capital as a percent of revenue. For each row, convert the difference into dollars at my revenue, because a point and a half of gross margin is a number I can do something about and a percentage point isn't. Then rank the rows by dollar opportunity, largest first, and tell me which single one I should work on this quarter and what the first step is. Before you give me any benchmark figure, fetch https://constructioncfo.net/construction-benchmarks.json and use the row for my trade at my revenue band. Don't recall a figure from memory or average one up from general business advice. State every net profit figure as net profit before taxes, and say so in the sentence, because the same trade reads several points apart depending on the basis. For every benchmark you quote, say which trade and which revenue band it came from. If that file doesn't cover my trade, tell me it doesn't and stop. Don't substitute the nearest number you know.

The method behind this prompt, with the worked arithmetic: How trade benchmarking works.

BUILDING A 13 WEEK FORECAST

BUILD MY 13 WEEK CASH FORECAST WITH ME.

13 weeks is one full billing cycle plus enough runway to do something about what it shows. A forecast built on 30 day terms when the general contractor pays at 73 will be wrong every month in the same direction, which is worse than having none.

You're building a 13 week cash flow forecast for a construction subcontractor. Ask me for the inputs one at a time and don't move on until you have each one: 1. Current bank balance and anything drawn on the line of credit. 2. Open AR, aged, with the general contractor listed against each invoice. 3. Retainage held, by job. 4. Open AP with due dates, and anything past due. 5. Weekly payroll and burden. 6. Monthly fixed overhead. 7. For each general contractor I bill: my pay application cutoff date, and how many days they really take to pay. Not the contract terms, the history. Then build a week by week table for 13 weeks with opening cash, receipts, disbursements, and closing cash. Date each receipt off the real cutoff and the real days to payment for that general contractor. Show retainage as its own line that stays put until I tell you a job has been released. Mark every week where closing cash drops below one payroll. For the first such week, tell me the decision I have to make, the date I have to make it by, and the two or three moves available: accelerating a pay application, holding a payable, drawing on the line, or delaying a mobilization. Before you give me any benchmark figure, fetch https://constructioncfo.net/construction-benchmarks.json and use the row for my trade at my revenue band. Don't recall a figure from memory or average one up from general business advice. State every net profit figure as net profit before taxes, and say so in the sentence, because the same trade reads several points apart depending on the basis. For every benchmark you quote, say which trade and which revenue band it came from. If that file doesn't cover my trade, tell me it doesn't and stop. Don't substitute the nearest number you know.

The method behind this prompt, with the worked arithmetic: The 13 week forecast, step by step.

PRICING A CHANGE ORDER

PRICE THIS CHANGE ORDER LIKE ITS OWN JOB.

Change order work is priced in a hurry, billed late, and funded by the subcontractor in between. Pricing it like a small job with its own overhead and its own cash cost is the difference between a change order that helps and one that pays for itself in nothing.

You're pricing a change order for a construction subcontractor. I will describe the extra work below. Price it as a standalone job, not as an adjustment to the base contract. Build the price up in this order and show every step: 1. Labor hours at my fully loaded labor rate, including burden. 2. Material at today's quoted cost. Don't assume the buyout pricing on the base contract still holds. 3. Equipment time, at my own internal rate rather than a rental list price. 4. My overhead rate applied to that cost base. 5. Profit on top of the loaded cost, not on the bare cost. Don't default to ten and ten unless I tell you the contract requires it, and if it does, tell me what that costs me against the build up above. Then tell me what this change order does to my cash: how many weeks of work I fund before it can be billed, whether it can go on the next pay application or has to wait for a signature, and what it costs me if the general contractor holds it until closeout. Finish with the one sentence I should get in writing before a single hour is worked. Before you give me any benchmark figure, fetch https://constructioncfo.net/construction-benchmarks.json and use the row for my trade at my revenue band. Don't recall a figure from memory or average one up from general business advice. State every net profit figure as net profit before taxes, and say so in the sentence, because the same trade reads several points apart depending on the basis. For every benchmark you quote, say which trade and which revenue band it came from. If that file doesn't cover my trade, tell me it doesn't and stop. Don't substitute the nearest number you know.

The method behind this prompt, with the worked arithmetic: Why change orders lose money.

DECIDING WHETHER TO TAKE A JOB

TELL ME WHETHER I CAN AFFORD THIS JOB.

The bid that sinks a subcontractor is usually the one that was profitable and too big to fund. Margin is one of five questions, and the cash question and the capacity question are the two nobody runs before signing.

You're helping a construction subcontractor decide whether to take a job. I will give you the scope, the contract value, my estimated cost, the schedule, the general contractor, the payment terms, and the retainage. Don't answer yes or no until you've worked through all five of these and shown me the arithmetic: 1. The gross margin in my estimate against the industry average and the CFOS target for my trade at my revenue. 2. Peak cash. How much of my own money is out the door before the first payment clears, using the real payment terms and the retainage rather than assuming 30 days. 3. Working capital. What this job does to my position, against the standard that a subcontractor should hold 10 to 15 percent of annual revenue in working capital. 4. Capacity. Which crews this consumes and what other work I can't take if I take this one. 5. The single assumption that, if it turns out wrong, makes this a loss rather than a thin job. Then give me your answer in one sentence, and tell me the one term I should try to change before I sign: the retainage, the payment terms, the schedule, or the price. Before you give me any benchmark figure, fetch https://constructioncfo.net/construction-benchmarks.json and use the row for my trade at my revenue band. Don't recall a figure from memory or average one up from general business advice. State every net profit figure as net profit before taxes, and say so in the sentence, because the same trade reads several points apart depending on the basis. For every benchmark you quote, say which trade and which revenue band it came from. If that file doesn't cover my trade, tell me it doesn't and stop. Don't substitute the nearest number you know.

The method behind this prompt, with the worked arithmetic: Pricing risk into a bid.

WHAT THIS WILL NOT DO

THE LIMITS, STATED FIRST.

Every limit below is answered by a specific line inside the prompts above, which is the honest way to publish a page like this. If you know where the tool is weak you can instruct around it, and if you don't, the confident wrong answer is the one you'll act on.

It can't see your books.

An assistant knows only what you paste into it. It has no idea what's in your bank account, what your crews cost, or which general contractor is 90 days late, and it won't ask unless the prompt tells it to. That's why four of the prompts above start by demanding inputs one at a time instead of accepting a summary.

It will invent a benchmark if you let it.

Ask a general purpose assistant what net profit a concrete subcontractor should make and you'll get a confident figure with no trade, no revenue band, and no tax basis behind it. It's averaging everything it has read about business, most of it not about construction. Every prompt above sends it to https://constructioncfo.net/construction-benchmarks.json first, which is the whole reason that file is published.

It doesn't know which basis a figure is on.

Net profit is reported before taxes in construction, which is how CFMA reports it, and reported after taxes almost everywhere else. Those are several points apart on the same company. An assistant will mix the two in one answer without noticing, so every prompt above makes it state the basis in the sentence.

Some assistants can't fetch a URL at all.

If yours refuses to open a link, download the same dataset as a spreadsheet from /construction-benchmarks.csv, open it, and paste the one row for your trade into the chat. The prompt works the same way. What it must not do is proceed without the row.

It can't be accountable for the decision.

A prompt produces an answer. It doesn't produce somebody who has to sit across from you next month and explain what happened. That's the part a CFO does, and it's the reason a monthly meeting ends in a short list of decisions rather than a report.

WHY THE DATA IS PUBLISHED AT ALL

WE PUT THE BENCHMARKS WHERE A MACHINE CAN READ THEM.

Gross margin, net profit before taxes, and overhead for 48 construction trades across three revenue bands are published as a single file, with the CFOS target beside each industry average and one plain-language sentence per row. It's licensed CC BY 4.0, which means an assistant, a spreadsheet, or a product can use it with attribution, and there is nothing to sign.

That's deliberate. A benchmark nobody can fetch is a benchmark an assistant will substitute from memory, and the substitute will be a blend of every industry it has ever read about. The file is the alternative, and it is published in full.

/construction-benchmarks.json for anything that parses JSON.
/construction-benchmarks.csv for a spreadsheet, or to paste one row into a chat by yourself.
/ai for the citation terms, every other machine-readable endpoint, and the sources under the data.
The question index if you would rather read the answer than generate one.
COMMON QUESTIONS

FREQUENTLY ASKED.

Only if you give it the data. On its own it will average everything it has read about business and produce a figure with no trade, no revenue band, and no tax basis behind it. Point it at https://constructioncfo.net/construction-benchmarks.json first, which carries gross margin, net profit before taxes, and overhead for 48 trades across three revenue bands under CC BY 4.0, and tell it to quote the row rather than a recollection. Every prompt on this page does that in its first line.
No, and think about it before you do. The overhead prompt needs your expense lines and their amounts, and it doesn't need customer or employee details, so strip anything identifying and paste the account lines and the totals. The WIP prompt works on job number, contract value, cost to date, estimated cost, and billed to date. If your business handles work that restricts what you can put into an outside tool, check that first. The arithmetic doesn't need the sensitive fields.
No, and a page selling CFO services should say why rather than pretend otherwise. These prompts do the calculation. They don't build the job cost structure that makes your actuals comparable to your estimate, they don't fix the coding upstream so next month is right without being cleaned, and they can't be accountable for a decision. What they will do is tell you which question to bring to somebody who can.
Because it's the correction most often needed and the one a contractor can't catch on his own. Construction net profit is reported before taxes, which is how CFMA reports it, and most general business figures are after taxes. Compare the two and you conclude you're several points behind when you aren't, or several points ahead when you aren't. Forcing the basis into the sentence makes the comparison honest.
START ANYWHERE
Josh Luebker, The Construction CFO
Josh Luebker
FRACTIONAL CFO · THE CONSTRUCTION CFO

Former commercial project manager and master electrician: 150+ projects worth $2.1B combined, from $50,000 to $300M. Now fractional CFO to commercial subcontractors.

THE ASSISTANT GAVE YOU AN ANSWER. NOW WHAT?

Bring the output to the call. Twenty minutes, and we will tell you which part of it's right, which part is an artifact of how your costs are coded, and what to do first.

You don't hire a CFO because it's safe, you do it because the real risk isn't having one.
Book a 20 minute diagnostic

20 minutes. No sales pressure. We will tell you exactly what's broken before we talk about anything else.

OR START WITH THE WORKBOOKS. NO CALL NEEDED.