Most small business owners are careful about the work and casual about the document that describes it. The site visit happens, a price gets scribbled on a card, and at eleven that night a quote goes out as three lines in an email. The job is won. Weeks later the margin has vanished into an extra coat of paint, a wall longer than remembered, and a favour nobody wrote down. All of that was decided at the quoting stage, not on site.
AI is genuinely useful for the typing part of quoting: turning scrappy site notes into an itemised document, applying your own rates consistently, and remembering the exclusions you always forget. It is not a source of prices, and it will smooth over exactly the ambiguity that costs you money. This guide builds the workflow so it does the first job and gets caught on the second. How these guides are made is on the About page.
Why quoting is where small businesses lose money
Getting a quote wrong rarely feels like a mistake. It feels like a busy month: invoices went out, money came in, and no report says “underpriced four jobs in March”. The loss is invisible because nothing went wrong in a way anyone notices.
It leaks in three places: scope you performed but never priced, which is unpaid work dressed as good service; hours estimated from memory at the end of a long day, where prep, the return visit and the tidying are quietly forgotten; and inconsistency, since a quote built from scratch each time tells you nothing about which jobs are worth having.
Underpricing is usually not a rate problem. Most owners know what their time is worth. It is a boundary problem — the rate was fine, the description of what it covers was not. “Repaint bathroom” is not a price; it is an opening offer in a negotiation the customer did not know they were in.
So AI helps here, but more narrowly than the vendors suggest. It will not tell you what to charge. It holds the document’s shape steady, applies your numbers the same way every time, and gives you something specific to correct — far easier than staring at a blank page trying to remember the last one.
What to give the AI before it can help
A model with no context does not write a quote. It writes a template with a number in it, and the number will be invented. Four inputs turn that into a draft you can work from, and all four are things you already know.
Your rates, as numbers rather than adjectives
Write your rate card down once, in a single file, in the plainest possible terms. Not “competitive hourly rates” but the figure. Include the awkward parts: the smallest billing increment, the minimum charge for a call-out, the markup you apply to materials, what you pay a subcontractor and whether you add anything on top, travel, and how tax is shown. If you bill a day rate, say what a day is.
This file is the most valuable part of the workflow, because it is the one thing the AI must never improvise. Every number in the finished quote should trace back to it. If you have never written it down, do that before anything else.
Scope, written as things you will physically do
Scope is a list of actions with quantities, not a description of an outcome. “Prepare walls” is an outcome; “fill nail holes and cracks in four walls, sand back, apply one mist coat and two finish coats to walls and ceiling” is scope. Only the second can be priced, argued about, and compared with what happened.
When you give the AI your scope, give it in that form: verb, object, quantity, and the surface or room it applies to. It will mirror your level of specificity. Vague input comes back vague, and vague is what customers read generously.
Exclusions and assumptions, listed as their own thing
Most owners carry their exclusions in their head: making good, waste removal, furniture moving, permits, parking, anything behind a fixed panel. Those are the items that turn a profitable job into an argument. Give the AI the list explicitly and tell it to reproduce it word for word, not paraphrase it into something friendlier.
The job notes, pasted rather than summarised
Paste the customer’s own message in their own words, your visit notes, and the constraints: access, parking, working hours, who else is on site, what moves and who moves it. Do not tidy it up first. The model spots a contradiction between “the tenant moves out Friday” and “three coats with drying time” faster than you will at nine in the evening.
A copy-ready prompt for a first-draft quote
This prompt assumes you have the four inputs above. Replace every bracket, delete any line that does not apply, and keep the structure: the rules about inventing information are doing most of the work.
You are drafting a quotation for a [BUSINESS TYPE] working in [REGION].
Use only the information below. Do not invent prices, rates, timeframes,
quantities or inclusions. If a detail you need is missing, write
[NEED: what is missing] instead of guessing.
--- MY RATES (use these exactly; do not round them) ---
Labour: [RATE] per person-hour, billed in [INCREMENT]
Minimum charge / call-out: [AMOUNT]
Materials: trade cost plus [PERCENT] markup
Subcontractor: [WHO] at [RATE]; add [PERCENT] handling or 0
Travel: [RULE]
Tax: [HOW IT IS SHOWN]
--- THE JOB ---
Customer's request, verbatim: [PASTE]
My site notes: [PASTE]
Access, parking, hours, constraints: [PASTE]
--- SCOPE: WHAT IS INCLUDED ---
[Verb + object + quantity + surface, one line each]
--- EXCLUDED UNLESS QUOTED SEPARATELY ---
[Reproduce this list exactly, do not paraphrase or soften it]
--- ASSUMPTIONS ---
[Site readiness, who supplies what, dates, drying or curing time]
--- OUTPUT, IN THIS ORDER ---
1. One sentence describing the job as priced.
2. Itemised scope, grouped by area, each line with quantity and unit.
3. Price breakdown showing LABOUR HOURS as well as money: hours x rate,
materials, subcontract, travel, then total.
4. Exclusions and assumptions, copied exactly from above.
5. Validity period: [DAYS].
6. Three questions I should ask the customer before this goes out.
Rules on style: plain and specific. No adjectives about quality.
No "we are pleased to offer", no "valued customer", no discounts,
no goodwill gestures, no item described as free or included at no
cost unless it is listed in my scope above. Keep sections 1-5 under
400 words. Do not write the covering email yet.Two details earn their place. Asking for hours as well as money shows the estimate you are actually committing to, which is where underpricing hides. The ban on discounts and goodwill gestures exists because models are relentlessly generous with other people’s labour.
Worked example: one bathroom, three drafts (illustrative only)
The figures below are invented to show the arithmetic and are not a guide to what anyone should charge in any trade or country. Your rates come from your rate card, not from this article.
What went in
A customer message asking for “the bathroom redecorated before we move in”. Site notes: four walls and a ceiling, a water stain above the shower, a radiator to remove and refit, tiles to protect, one area of blown plaster, a two-week window. The rate card gave a labour rate, a materials markup, and a rule that a second visit is charged as a return call-out, not absorbed.
What came back
A neat document: six hours of labour, a materials figure, a correctly copied exclusions list, a tidy total. It read well, and in three specific ways it was a promise to work for less than the job required.
What the owner changed
First, hours. Six covered the painting and nothing else. The true figure was closer to eleven once filling and sanding, protecting the tiles, removing and refitting the radiator, and the second visit for the sealer coat over the stain were counted. Painting two coats is not two coats of time.
Second, an inclusion the owner never wrote: the draft had absorbed the blown plaster into “preparation”, because in its experience of renovation writing plaster repairs usually sit there. That is unpaid work with a friendly name.
Third, a missing line. The owner always includes waste removal, and the draft had left it out of scope and out of the price. It was added back as a priced line rather than a promise.
The corrected quote was longer, less charming and roughly a third higher. It was also the one the owner could afford to do, which is the only test that matters.
Reviewing the draft for scope creep and underpricing
Read the draft twice with a different question each time. First pass: what have I just promised to do? Second: what have I promised to do it for? Leave tone until both are done, because a well-written quote is harder to interrogate than a clumsy one.
The words that create scope creep
Scan for vague nouns and totalising adjectives: prepare, make good, tidy up, complete, all, throughout, as discussed, including, necessary works, and anything ending in “etc”. Each one is a hole a customer can reasonably push through. “Complete redecoration” means whatever the customer pictured. Replace every one with a countable action, or move it to exclusions.
Where the hours disappear
Check the labour figure against the scope line by line, not against your gut. Prep takes longer than finish, second visits exist, and drying time is calendar time you cannot compress by working faster. Anything you collect, hire, dispose of or return is hours. A single round number of hours for the whole job is a placeholder, not an estimate.
The generosity problem
Models add value you did not ask for: a line including the door frames, an offer to touch up the hallway, a discount to make the total friendlier. Each is money out of your pocket and a precedent for the next conversation. Delete them. If it is not in your scope list it is not in the quote; the only acceptable version is a priced optional extra.
The pre-send checklist
Five minutes here is cheaper than an afternoon of unpaid work.
| Check | Why it matters |
|---|---|
| Every price traces to my rate card | A number the AI invented is a number you have to honour |
| Labour hours match the scope, line by line | Underpricing hides in a single total figure |
| No vague words left in scope | “Prepare” and “make good” are open-ended |
| Exclusions copied exactly, not paraphrased | Softer wording reads as an inclusion |
| Assumptions stated as facts the customer confirms | Moves the risk of a wrong assumption off you |
| Second visits, drying time and disposal priced | The three most commonly forgotten hours |
| Optional extras priced separately, never “included” | Keeps the base price honest |
| Validity period and payment terms present | A quote with no expiry is an option you granted for free |
| Covering email written by you, not the model | The relationship part should sound like a person |
| Nothing in it is advice you are not qualified to give | Describe the work, not the regulations around it |
What should stay a person’s job
Rates, terms, and the judgement about whether a job is worth having. No model knows your market, your backlog or the customer in front of you, and none can price risk. The document is the output of your thinking, not a replacement for it.
Plainly, what this guide is not: a workflow for producing a document, not financial, legal, tax or contractual advice. Payment terms, liability, and what your jurisdiction requires in a consumer quote are matters for you and your advisers, and belong settled before a template goes near a customer. The standards we hold ourselves to are in our editorial policy, and the tools section covers the software side of running a small business.
Frequently asked questions
Can I let AI set my prices?
No, and the attempt is the fastest way to lose money with this workflow. Models have no access to your costs, your overheads, your local market or your capacity this month, and any figure they offer is a plausible-looking invention. Give it your rates and it will do the arithmetic and the formatting correctly, which is the useful part. Ask it for rates and you get a confident guess dressed as a benchmark.
Will customers be able to tell the quote was AI-assisted?
Usually not from the document, since a quote is a list of work and money. What customers notice is vagueness about what is included, or a price that changes after work starts. Clear scope reads as professional whoever typed it. Write the covering email yourself; that is the part with a voice in it.
What if the AI writes something that is not in my scope?
Delete it. This happens constantly, and it is the single most expensive failure mode in the whole process, because the draft reads better with the extra generosity in it. Treat every addition as a claim you will have to honour. If the work genuinely should be included, add it to your scope list and price it. If it should not, it belongs in exclusions.
Should I paste customer details into a chatbot?
Use the minimum you need. A quote rarely needs more than the job description, property details and constraints, so strip names, contact details and anything sensitive first, and check whether your provider uses inputs for training. If a message is confidential, describe the job in your own words instead of pasting it.
Start with one real job this week: build the rate card, paste the notes, run the prompt, then spend your time on the two passes that matter — what you promised, and what you promised it for.
Next step: if the bottleneck is the messages coming in rather than the quotes going out, start with customer replies that do not sound like a robot. If you are still choosing tools, the directory lists what each one is actually good at.
Want the wording ready-made? The ChatGPT prompt library has copy-ready prompts for quoting, chasing and client updates, organised by job.
Last reviewed and updated on November 4, 2026.