If you run a trades or services business, you know where the bottleneck is. It is not the work. It is the paperwork before the work. Somebody walks a job, takes notes on a phone, then loses an evening turning those notes into a quote professional enough to send. Meanwhile the customer got two other estimates on Tuesday and has stopped waiting.
AI is genuinely good at that gap. Not at deciding what to charge, but at the writing, formatting, and checking between a site visit and a document a customer can say yes to. Used carefully it turns a two hour evening job into a twenty minute review. Used carelessly it produces a beautiful quote with a margin you cannot survive.
Where AI Genuinely Helps On A Quote
Everything on this list has the same shape. The AI works with words and structure, not judgment about money.
- Drafting scope language. Feed it your rough site notes and ask for a clean scope of work in your standard format. Three lines of shorthand become properly worded line items with inclusions and exclusions. You edit rather than compose.
- Reformatting past quotes. Ask it to pull scope language from three similar past jobs and adapt it to this one. Your best writing gets reused, not rewritten.
- Catching missing line items. The underrated one. Give it your checklist for the job type plus your draft, and ask what is on the checklist but not the quote. Permits, disposal, restoration, and travel are the usual escapees, and every one comes out of your margin.
- Writing the cover note. The paragraph explaining why you scoped it this way and what happens next. Most estimators skip it, and it is often what wins the job.
- Proofreading against the source. Ask it to compare the finished quote back to your site notes and flag anything that never made it onto the document.
Faster is the point, and it is worth real money when quotes sit in a pile on Thursday night. Same argument we make in why saving time takes priority over saving money.
Where AI Must Not Be Allowed To Decide
Here is the line, and it is bright. AI drafts. Humans decide anything that costs money, creates an obligation, or gets signed.
- Pricing. Your prices come from your cost data, your local market, your capacity this month, and your read of the customer. A language model has none of that. It has a sense of what numbers look plausible, which is a terrible basis for a bid.
- Margin. Never let a tool decide markup. That number is a strategy decision and it belongs to whoever owns the profit and loss.
- Feasibility. Whether a job can be done that way, in that building, with that crew is a question for someone who has been on site. AI will happily produce a confident scope for something physically impossible.
- Code, permits, and safety. Requirements vary by jurisdiction and change over time. Use AI to draft the question you ask your inspector. Do not use it as the answer.
- Anything a human signs. Terms, warranty language, liability, payment schedules. A person with authority reads it and owns it.
The reason for that line is documented. NIST’s Generative Artificial Intelligence Profile, published in July 2024, defines confabulation as “a phenomenon in which GAI systems generate and confidently present erroneous or false content.” Confident and wrong is the failure mode, and a quote is exactly the document where confident and wrong sails through review.
In a 2025 post on its technology blog, the Federal Trade Commission wrote that “firms deploying these AI systems and tools have an obligation to abide by existing laws.” The quote goes out under your name and your license number.
Treat It Like A Junior Staff Member
Harrison’s framing is the one we keep coming back to: treat AI like a junior staff member. Not a calculator, not an oracle. A fast, eager new hire who writes well, has read a great deal, has never seen your business, and will never tell you they are unsure.
Think about how you would manage that person. You would give them your templates and past quotes, let them prepare drafts and flag questions, and absolutely not let them set prices in month one. You would check their work until they earned the right to be checked less.
That framing also protects you from the trap NIST names. The same publication warns that “humans may over-rely on GAI systems or may unjustifiably perceive GAI content to be of higher quality than that produced by other sources,” which it identifies as automation bias. A polished document reads as more correct than a messy one, regardless of whether it is. NIST recommends organizations “establish policies to define and differentiate roles and responsibilities for human-AI configurations and oversight.” For a twelve person contractor that is one page: what the tool drafts, what a human approves, and whose name is on the approval.
Check Every Number Against Your Own Cost Data
Even when AI is not setting prices, numbers leak into drafts. It copies a quantity from the wrong past job, or adds a line item at a plausible rate. Language models are not calculators.
- Keep pricing in a system that does math. Your estimating software or price book holds the rates and does the arithmetic. AI touches the words around the numbers, not the numbers.
- Verify quantities against the site record. Every measurement should trace back to a photo or a walkthrough note. If it cannot be traced, it does not go on the document.
- Sanity check the total. Compare against your typical range for that job type and size. If it is well off your normal, find out why before it goes out.
- Track win rate and as built cost. If quotes are getting faster and margins are getting worse, the process is broken no matter how good the documents look.
A Template Plus AI Beats AI Alone
The biggest improvement most businesses can make has nothing to do with AI. It is a real template: a fixed quote structure, a standard set of inclusions and exclusions, a checklist per job type, and boilerplate terms your attorney has reviewed.
Give an AI a blank page and it invents a structure, slightly differently every time. Give it your template and it fills your structure, in your language, with your exclusions. Output becomes consistent and reviewable fast, because the reviewer knows where to look. Build the template first even if you never adopt AI. Teams that do it in that order get a durable improvement. Teams that skip it get faster inconsistency, which we cover in how to actually prepare your team for AI without the hype.
The Bottom Line
AI will not estimate your jobs. It will help you get estimates out faster, written better, and with fewer forgotten line items. Speed to quote wins work. Accuracy of quote keeps the work profitable. Two different problems, and AI is a real help with the first and a real hazard if you let it near the second.
Draw the line in writing. Templates and cost data hold the numbers. AI drafts the language. A named human reviews every quote and can explain every line on it. Confirm current features and pricing with any vendor before you commit, but that division of labor will still be right long after today’s products are gone.
If you want help setting this up, from the quote template to the checklists to the guardrails on what the tool may and may not touch, that is work we do. We serve small and mid sized businesses across Denton County and North Texas, and we will be straight about which parts of estimating AI should stay out of. Contact us today
Sources:
Comments are closed