Verification pending. This write-up describes what was built and how it works. It contains no performance figures, and it will be updated with measured results once they are verified.
Challenge
In residential roofing, the homeowner wants a number at the kitchen table. The salesperson wants to give one. What usually happens is a promise to “send it over tonight,” an evening of estimating, and a proposal that arrives after a competitor’s.
Existing workflow
Measurement report in one window, spreadsheet in another, a word-processor template in a third — and pricing that depended on who built the proposal.
What we built
- A proposal generator that takes roof measurements and a chosen scope, applies the price book and waste factors, and produces a branded PDF with good / better / best options.
- A quote tool for permanent lighting that traces eaves and rakes on satellite imagery, corrects rake lengths for roof pitch, prices per linear foot and outputs a one-page PDF plus structured data. It can also read KML/KMZ files and measurement-report PDFs.
- A path to payment in the Certified AI operating system (on staging): proposal → agreement → e-signature → payment link → a person confirms settlement → onboarding.
Human involvement
Prices come only from the owner’s price book, and every proposal is reviewed before it’s sent. Payments are recorded as settled only after a person confirms them.
Result
Both generators are built and tested. We aren’t publishing speed or close-rate numbers because we haven’t measured them against a baseline.
Lessons learned
- Verify vendor APIs before you build. One e-signature vendor’s API was enterprise-only; another’s wasn’t actually enabled on the account.
- “Completed” isn’t “paid.” Bank payments pass through a processing state before they settle; the system models that explicitly.
- Structured output beats a pretty PDF alone. Saving every quote as data makes follow-up and reporting possible.