Case study · GEC Roofing & Restoration
We ran the experiment on a real roofing company first.
Before Trailpoint recommended AI to anyone, we built it into GEC — a working roofing and restoration company we already consulted for, where one of our founders helps run business development and estimating. Here's what actually paid off: 4 hours saved per insurance supplement, $5,200 in average added scope per supplement, across 50+ live jobs.
The situation
GEC runs roofing, gutter, siding, and repair work — estimate-driven, insurance-heavy, deadline-pressured. The two workflows that decided profitability were the two that ate the most skilled time: building accurate estimates fast enough to win the job, and documenting insurance supplements thoroughly enough to get paid for all the work actually done.
Supplements were the expensive one. Done right, a supplement means documenting every legitimate item the initial insurance scope missed. Done under deadline pressure with a manual process, items get missed — and missed items are revenue that was already earned and never collected. Estimating had the same shape: slow enough to cost jobs, manual enough to vary by who built it.
What we built
Engines for the two workflows that move money.
The insurance-supplement engine
Reads the job file and the carrier scope, drafts the supplement with every documentable item, and hands it to a human for review. The 4-hours-per-supplement saving isn't from typing faster — it's from not rebuilding the same document from scratch every time. The $5,200 average in added scope is what thorough looks like when thorough is cheap.
The estimator engines
Four of them, all operational on live jobs: roofing estimates, gutter estimates, siding estimates, and repair estimates. Each one runs inside the AI tool the team already uses — Claude, in GEC's case — and is integrated with the systems GEC already runs: JobNimbus for jobs and contacts, CompanyCam for field photos and inspection notes, and live supplier pricing and availability. Every engine is hardcoded with GEC's own labor rates by trade and crew, so an estimate reflects what the job actually costs GEC to build, not a template's guess.
Working with an estimate is a conversation, not a spreadsheet session. Change a pitch, swap a shingle line, add a downspout run — you say it in plain English, the same way you'd chat with any AI tool today, and the estimate updates with real pricing and real labor rates behind it.
That integration detail is the difference between a demo and a tool. Software that doesn't know your labor rates, your suppliers, and your job files makes work; software that does, removes it.
Measured results
From live jobs. Not a pilot, not a projection.
Saved per supplement
Skilled estimating time back on estimates, not paperwork.
Average added scope per supplement
Documented, submitted, and approved work that a rushed manual process was leaving behind.
Live jobs run
Through the supplement engine — with 100+ more jobs run through the four estimator engines, inside a real operating business.
What it took
The honest part.
None of this worked on day one. The real obstacles: messy historical job data, long stretches of rate-book cleanup, and the unglamorous work of getting outputs consistent enough to trust on a live job. The first build of each engine took real time. Every one since has gone faster, because the lessons carried over.
What made it stick: every engine was built using the daily workflows GEC already ran, reviewed by the people who own the number it moves, and measured on live jobs before anyone called it done. That's the same standard we hold when the business isn't ours.
Your business next
The audit applies this same survey to your operation.
Two weeks. A ranked list of 5–10 places AI pays off in your business, each with a dollar estimate. You keep the roadmap either way.