Construction · 4 min read · October 10, 2026
Before an AI drafts a single roofing quote, read the ones you already sent. Here is what A&J Professional Services' own estimate history showed, how we are joining it to supplier orders, and why nothing drafts quotes for customers yet.
Start with the estimates you already sent, not with an AI. Pull your past estimates and line items, join them to the supplier orders behind each job, write the owner's pricing rules down, and only then let a system draft a quote that a person approves. If your history shows one markup on every job, fix the costing first.
Christian, who runs A&J Professional Services, wants quotes out faster. Every roofer does. The obvious build is an AI that reads the measurements and writes the quote. We did not start there, because a quoting system can only be as good as the pricing it learns from, and nobody had checked that pricing yet.
So the first job was reading. This article is what the reading found, step by step, so you can do the same with your own data before you pay anyone to automate it.
A&J had 485 estimates with 984 line items in its CRM. That history is the starting point for any quoting system, because it shows how the company actually prices, not how it thinks it prices.
A&J keeps its jobs in JobNimbus, a roofing CRM. We copied its estimates into a database for the build: 485 estimates with 984 line items. We also learned where proposals actually go out from. Only one of the 485 showed up in the office's sent email, because proposals are built and sent from inside JobNimbus. Any automation has to land there, not in somebody's inbox.
The same pull showed the gap. The places where pricing rules and pricing examples should live were empty. The owner's judgment about price was in his head and in old quotes, nowhere else.
When the margin is identical on small and large jobs, the estimates were built from a template, not from what each job cost. A system trained on that would repeat the template.
When we grouped the estimates by size, the margin came out the same on small jobs and large ones. Real job costs do not behave like that. A tear-off on a steep roof and a small repair do not cost the same share of the price. That pattern means the estimates were built from a fixed markup template, not from what each job actually cost.
That is not a criticism; most small contractors price that way. But it matters for automation. A system trained on that history would learn the template and repeat it, faster. You would get quicker quotes with the same blind spot.
None of the 485 estimates could be matched to a supplier order at first, because nothing connected the two. Pulling the supplier's order history is the step that makes real job costs visible.
The real cost of a roofing job shows up in the supplier order. At first, none of the 485 estimates could be matched to a supplier order, because nothing connected A&J's CRM to its ABC Supply account. On October 10, 2026 we pulled the supplier history: 323 closed orders and 65 open ones. Matching those to estimates is the step underway now.
Once an estimate and its order sit side by side, you can see what a job actually cost in materials against what was quoted. That is the number a quoting system should learn from.
In parallel, we wrote Christian's pricing judgment into a short quoting guide in October 2026: how he prices a roof by the square, how siding tiers differ, and what changes a price. It is a document, not software, and that is deliberate. A rule a person can read and correct is the base a system can follow later.
If you do one thing after reading this, do this one. Sit with whoever prices your jobs and write their rules down in plain words. It is useful even if you never automate anything.
As of October 10, 2026, no system drafts quotes for A&J's customers. A material-quote step exists and passes its tests on made-up data only, so it is not proven. We also tested pulling the text out of six real sent proposals, to see whether old proposals can be read reliably and used as examples later.
The order from here: match estimates to supplier orders, check the pricing guide against those matched jobs, then let a system draft a quote that a person approves inside JobNimbus. Each step is judged by replaying jobs the company already won. The same approach is in why a price list per trade comes before an AI estimate.
Common questions
It can draft one from measurements, material prices and your pricing rules. The quantities and dollars should come from plain code, and a person should approve every quote before a customer sees it. The hard part is not the drafting; it is having pricing rules worth following.
Your past estimates with their line items, the supplier orders for the jobs you won, and the owner's pricing judgment written down. If your estimates and your supplier orders can not be matched to the same job, start there.
A&J's proposals are already built and sent from JobNimbus, so we work around it instead of replacing it. Any automated draft should land back in the CRM the office already uses, where a person reviews and sends it.
The drafting is quick to build. Getting the history clean enough to trust takes longer, and it depends on how well past estimates and supplier orders line up. Plan the data work first and judge the system by replaying jobs you already won.
Sources
Next step
Book a call and bring an export of your last year of estimates. We will look at it with you and show whether your history is ready to automate, or what to fix first.
Written by
Chris Moreno builds custom AI systems for business operations. His writing draws on the work behind these systems: intake, follow-up, document workflows, and the checks that keep people in control.
Published October 10, 2026 · Connect on LinkedIn
Our methodology
The approach behind this work follows the four phases of Flo OS, our operating methodology for turning messy business workflows into systems that run cleanly and compound over time.
See how we work →Related reading
The estimating math for a renovation contractor passed its tests, and it still could not produce a number anyone should sign. The reason was underneath it: the price list. Here is what we found when we checked each input against a real estimate.
In a design-build company the time goes to waiting: measurements, photos and prices sit until one person gets to them. Here is the chain we are building for a contractor, stage by stage, and how each step gets tested.
Everything we built for A&J Professional Services, a roofing company in New Jersey, in the order we built it: the receptionist, the ad cleanup, the website and reviews, the content engine and the quoting work. What is live, what is not, and what the owner says happened to the business.
