Every good contractor has a toolbelt. This is mine - open to the trade.
Line items, field references, example photos, SOPs, checklists, and lessons learned - take what is useful and add it to your own toolbelt. The goal here is simple: a working database of real information that is accessible to anyone in the trade. No paywall. No subscription. No gatekeeping. Too much of what you need to do this job right is locked inside someone's head, buried in proprietary software, or sitting behind a login screen that costs you monthly just to access your own industry's knowledge. Roofing is dirty and it is hard work. The paperwork and the reference material that backs it up should not be.
Documents, templates, example photos, Xactimate line items, manufacturer spec sheets, installation references, and more - all pulled from the same database that powers the Field Desk.
I don't chase fluff supplements. I don't nickel-and-dime carriers over line items that don't matter. What I do is make sure every job gets built right - and that every line item required to build it right is on the scope.
The foundation is simple: a roofing system, a siding system, a gutter system - each one is a single continuous system when installed properly. No single component exists in isolation. Every piece is fastened, adhered, overlapped, or interlocked with another piece. You can't touch one part without affecting the parts connected to it.
Every job we write carries two non-negotiable commitments: a full manufacturer product warranty and our workmanship warranty. To honestly send a final invoice, we have to confirm that every product was installed per manufacturer requirements and everything meets local and state code. That is not a sales pitch - that is the contract.
So when a carrier approves a roof replacement but denies the flashings that have to be removed to install it, or approves a siding repair but ignores the house wrap that gets torn during detachment - those are not optional add-ons. Those are required components of the approved work. Code says so. The manufacturer says so. Our warranty says so.
Before we perform any partial repair, it has to pass a feasibility evaluation. Can we actually do this repair and still meet code, match existing materials, maintain manufacturer specs, and stand behind the work? If the answer is no - and on the majority of partial repairs, it is no - then we document why, communicate it clearly to the homeowner and the carrier, and request the scope that allows us to do the job correctly.
If the carrier still directs a repair we cannot warrant, we do not refuse to help the homeowner. We explain exactly what the limitations are, put it in writing, and get a waiver signed. The waiver is our way of documenting that everything was communicated clearly - that the property owner understands an incomplete scope or partial repair may void manufacturer warranties, may not meet local code requirements, and would not carry our workmanship warranty. We do not use it as a pressure tactic. We use it because the homeowner deserves to know exactly what they are agreeing to before work starts.
What usually happens is straightforward. When the property owner understands the options and realizes what they would be signing away, they do not want to sign it. And that puts them in a very strong position to go back to their carrier and receive a properly written scope. At the end of the day, it is their claim - not mine. My job is to make sure they have the information and documentation to get the work done right.
That is the whole strategy. Build it right, document everything, and let the code and the manufacturer specs do the talking.
Code and manufacturer compliance review for ensuing damages during covered roof replacements.
Feasibility and compliance verification testing required before any partial roof repair.
Code and manufacturer compliance review for partial siding repair feasibility.
Notes from building tools and learning AI inside the trade.
I stopped traveling, came home, and took an office job. Staring at line items, spreadsheets, price lists, and scope of loss documents all day. Dealing with adjusters, salesmen, and the ops team from a desk instead of a roof. Being home every night at a normal time was great - but there was nothing to build. Nothing to obsess over. That gap did not last long.
I figured out fast that you cannot communicate, access, or process information in an office environment the same way you can in the field. The pace is different. The inputs are different. So I started playing with AI - testing every LLM, automation, and tool I could find to see what could actually keep up with the volume and complexity of production estimating and supplementing.
Most of 2024 and into 2025, I lived in ChatGPT. I built custom GPTs, ran Zapier and Make.com automations, and started learning my way around Supabase and basic database structure. It handled estimating, supplement drafting, email writing, and information lookup well enough to prove the concept. But it also showed me where the ceiling was.
Early 2026, several things hit at once. Claude 4.6 dropped. The OpenAI situation blew up. I got introduced to Claude Code. I went to the Orlando Builders Show, RoofCon, and sat in on every AI class, workshop, and seminar I could find. I have spent an embarrassing amount of money on online courses, mentors, and self-proclaimed AI experts. I have also downloaded and tested what feels like every new AI product that hits my feed or my DMs.
Here is what I keep coming back to: the data is everything. If your data is organized, labeled, factual, and consistent - everything you build on top of it is just features. Estimating tool, supplement processor, email automation, voice agent, onboarding guide, training system - it does not matter what you are building. If the foundation is solid, you can put whatever structure you want on top of it.
I understand building. Roofers, siders, gutter guys - we all know how to put components together once the foundation is right. That framing is what finally made this click for me.
So I stopped chasing tools and started building a foundation. Google Drive first - folder structure, SOPs, checklists, templates, everything I use daily. Connected it to ChatGPT and Claude. Built Claude Skills and custom GPTs around specific daily tasks: scan a Hover and produce a material, labor, and margin estimate; scan a scope and generate a pre-supplement or turnover request against 600-plus Xactimate line items with code and manufacturer-backed F9 notes; pull photos and CRM threads to build change orders or PWI supplements; draft emails from pre-built templates. That alone killed most of the daily monotony and kept me focused on the edge cases and exceptions.
Then I started building actual apps. Crashed a few things in Lovable learning how vibe coding works. Moved into Claude Code and Netlify. Dialed in a stack that made sense:
Every time a Hover gets uploaded, a scope gets scanned, a photo gets dropped in, or a question gets asked - it pulls from the same database. How to read a Hover. How to analyze a scope. How to match a photo to a similar loss in the library. How to answer an adjuster who says you can detach and reset that twenty-year-old skylight. Factual, consistent, real-time answers built on real field knowledge.
The foundation is built. Most of the day-to-day now is iteration - cleaner data, sharper prompts, fewer manual steps.
No pitch. No pressure. Just a straight answer about what your property needs.