The size of the prize, and why claims keep dying
The U.S. storm restoration market — defined as residential and small-commercial roofing work paid through insurance claims — sits at roughly $50-65 billion in annual GMV depending on whose numbers you trust (TRA, IBHS, IRMI). It's growing. Hail events are getting more frequent. Insurance carriers are getting tighter on payouts. And the gap between "claim approved" and "claim paid in full" is where most contractor revenue actually leaks.
Here's what kills claims, in rough order of frequency:
The pattern is consistent: the contractors making real money in storm work aren't the ones with the lowest bids or the prettiest websites. They're the ones with the best documentation discipline. Better photos, cleaner Xactimate-friendly write-ups, faster supplement turnaround, organized communication trail.
Documentation discipline used to be a tradeoff. You either had a senior office person babysitting every claim, or you accepted a 25-35% leakage on what insurance owed you. AI changes that math.
The contractors winning storm season in 2026 aren't the ones with the most door-knockers. They're the ones whose documentation is so airtight the adjuster has nothing to deny.
The line: where AI is allowed (and where it isn't)
Before we get into the workflow, the legal piece. Insurance work is regulated at the state level, not the federal level, and the rules vary in ways that matter.
Important · not legal advice
Public adjusting laws differ by state. Texas, Florida, California, New York, and others require licensing for anyone negotiating claims on behalf of a homeowner for a fee. AI cannot be a public adjuster. Your contractor relationship with the homeowner is a separate thing — you're documenting your work, not representing them in negotiation. Stay on your side of the line. When in doubt, your public adjuster relationship handles the negotiation; you handle the roof.
What AI can do (with appropriate human oversight)
- Document damage with timestamped, geotagged photos and structured captions
- Generate Xactimate-friendly line item descriptions and scope language
- Calculate RCV, ACV, depreciation, recoverable depreciation
- Draft supplement letters for human review and signature
- Organize photo packages by elevation, by damage type, by claim section
- Draft homeowner-facing explanations of claim concepts (deductible, RCV vs ACV, depreciation)
- Track claim status and timeline, alert when a deadline approaches
What AI should not do
- Negotiate directly with insurance carriers on behalf of a homeowner
- Sign supplement letters or claim correspondence
- Speak (voice) directly to adjusters as if it were a human representative
- Make legal claim representations (e.g., "this damage is covered under your policy")
- Draft policy interpretations without licensed review
The clean rule: AI prepares; humans negotiate. Configure your workflow so every outbound carrier-facing communication has a human signature and human accountability.
The claim journey, stage by stage
Here's the full sequence a storm-restoration claim moves through, with what AI can do at each step.
FNOL — First Notice of Loss
Homeowner notices damage, calls insurance to start a claim. Before that call happens, AI can: collect initial damage details, run a preliminary feasibility check ("is this likely a covered claim?"), explain the FNOL process to the homeowner, and document the timeline starting from the moment of damage discovery.
What AI does: structures the homeowner's initial damage report, captures relevant photos, builds the timeline document. Keeps a clean record before any adjuster gets involved.
Pre-adjuster damage documentation
Before the insurance adjuster arrives, you should already have full documentation in hand. This is where AI documentation pays off the most — every photo categorized, captioned, geotagged, organized by elevation. Drone footage stitched into a complete roof tour. Interior damage documented with date stamps.
What AI does: ingests photos from drone, ground, and interior. Produces an organized claim packet with damage map, line items, and Xactimate-friendly descriptions. Time saved per claim: 2-4 hours.
Adjuster meeting
The adjuster shows up, walks the roof, and writes their initial estimate. This meeting is human-led — you, your team lead, or your public adjuster. AI should not be the one in the meeting. But it should have prepared everyone for it.
What AI does (before): generates a meeting prep doc — likely scope items, common adjuster pushback patterns from this carrier, photo references, line item supports, your target RCV. During: nothing — humans only. After: immediate documentation of what was agreed, what was disputed, action items.
Estimate review & gap analysis
Adjuster sends their initial estimate (usually as a PDF or Xactimate export). This is where the most money is left on the table. AI can ingest the adjuster's estimate, compare line-by-line against your scope, and flag every missing or under-priced item with photo evidence and reasoning.
What AI does: reads the adjuster's estimate, runs the comparison, generates a "gap analysis" doc — every line item missing, every quantity reduction, every depreciation override. Each flagged item references the specific photo and line of evidence. Time saved: 3-6 hours per claim.
Supplement letter drafting
Once the gap is identified, you need a written supplement request — a formal letter to the adjuster requesting additional line items with evidence. AI drafts it. Human reviews, edits, signs, and sends.
What AI does: produces a supplement letter draft with: cover narrative, itemized list of requested adjustments, photo references for each, scope/measurement justifications, and Xactimate-formatted line items. The contractor or public adjuster reads, edits, and sends. Drafting time drops from 1.5 hours to 8 minutes.
Carrier negotiation
The back-and-forth with the adjuster. Phone calls, follow-up emails, possibly a re-inspection. This is where licensed humans (your team or your public adjuster) handle the work directly. AI does not negotiate.
What AI does: tracks every interaction in a structured timeline, alerts on deadlines (most carriers have 15-30 day response windows), flags when a re-inspection is overdue. The conversations are human; the record is automated.
Re-inspection prep & supplement updates
If a re-inspection is ordered, AI assembles the brief: what was disputed, what new evidence supports the supplement, photo highlights to walk through. Whoever attends is fully prepared.
What AI does: generates the re-inspection prep document, updated supplement request, and revised gap analysis. Human attends the inspection.
Payment tracking & depreciation recovery
Initial payment is usually ACV (actual cash value) — RCV minus depreciation. After work is done, the homeowner can claim the recoverable depreciation by submitting proof of completion. Most homeowners forget. AI tracks the timeline and prompts the homeowner to file for recoverable depreciation, with the documentation pre-assembled.
What AI does: tracks every claim's payment status, alerts the homeowner when recoverable depreciation is filable, drafts the homeowner-side request with completion documentation. Average recovery rate goes from ~60% to ~95% on tracked claims.
Endorsement and claim closure
Final paperwork. Lien releases. Warranty registration. Customer satisfaction follow-up. AI runs all of it, hands off to humans for any signature requirements.
What AI does: manages the end-to-end paperwork checklist, queues homeowner signatures, registers warranties, triggers the review request workflow. Closes the loop.
The full storm system
This claim-side workflow pairs with the front-of-funnel storm activator system covered in The Roofer's AI Playbook — when hail or hurricane radar pings, AI builds the prioritized call list of past customers in affected ZIPs. Combined with claim documentation automation, you get end-to-end storm coverage from "radar trips" to "depreciation recovered."
Xactimate-friendly documentation: the format that wins
Adjusters work in Xactimate. Your documentation should map cleanly to it. Here's what AI produces on a well-configured deployment:
Photo packages
Every job photo gets:
- Date and timestamp (auto)
- GPS coordinates (auto)
- Elevation tag (front, rear, left, right) — AI inferred
- Damage type tag (hail strike, wind damage, mechanical, foreign object) — AI inferred
- Reference grid coordinates matching your roof diagram
- Caption referencing the matching Xactimate line item
Output: a single PDF (or shared cloud folder) with photos organized by elevation, damage type, and Xactimate line. Adjuster opens it and the work is done for them.
Scope documents
AI produces scope language that matches Xactimate's standard items — same units (squares, linear feet, each), same product specifications (manufacturer + line + style), same trade categories. The supplement letter doesn't ask for "extra flashing"; it asks for "RFG XGL3 — Drip edge — Replace, 88 LF — refer Photo 12, Photo 14, Photo 17."
Line item justification
Every requested line item carries: photo evidence, measurement source (Hover/EagleView/manual), code reference if applicable, and a one-sentence narrative. Adjusters approve faster when the work is done for them.
State-by-state: regulation snapshot
Public adjusting laws vary. Here's the high-level for the major storm-restoration markets — but consult your state insurance department or a licensed public adjuster for actual compliance advice:
| State | Public adjuster license required to negotiate? | Contractor "public adjusting" allowed? |
|---|---|---|
| Texas | Yes (TDI-licensed) | No — strict separation. Contractors document their own work; cannot negotiate claims for fee. |
| Florida | Yes (FL DFS-licensed) | No — actively enforced. AOB reform statutes apply. |
| California | Yes (CDI-licensed) | No — clear separation required. |
| Arizona | Yes (ADOI-licensed) | No — separation enforced. |
| Oklahoma | Yes (OID-licensed) | No — clear separation. |
| Georgia | Yes | No — public adjuster license required for fee-based representation. |
| Tennessee | No state PA license · light regulation | Looser, but homeowner-direct contracts still recommended. |
The takeaway: nearly every storm-belt state requires a licensed public adjuster for fee-based claim negotiation. Contractors document, build evidence, and submit work-side documentation. Negotiation is a separate licensed function. AI helps both sides — it does not replace either.
Real numbers: what storm-belt contractors are actually seeing
Here are anonymized numbers from three storm-restoration shops we've worked with — one Houston, one Tampa, one Phoenix. All ran on JobNimbus + Hover + their own internal documentation process. We deployed claim-documentation AI on top.
Shop A · Houston-area · ~80 claims/mo
- Documentation time per claim (pre-deployment): 4.5 hours · (post-deployment): 1.1 hours
- Supplement approval rate: 61% → 84%
- Average claim RCV captured (vs adjuster initial): 76% → 94%
- Time-to-payment: 41 days → 27 days
- Annual revenue impact: roughly +$840k from supplement recovery + +$220k from time savings reinvested in volume
Shop B · Tampa · ~50 claims/mo
- Documentation time per claim: 5.2 hours → 1.4 hours
- Recoverable depreciation collection: 58% → 92% of eligible claims
- Annual revenue impact: roughly +$390k from depreciation recovery alone
Shop C · Phoenix · ~30 claims/mo (smaller, hail-driven)
- Documentation time per claim: 3.8 hours → 1.0 hour
- Supplement approval rate: 55% → 79%
- Annual revenue impact: roughly +$210k
The pattern across all three: documentation time drops 70-80%, supplement approval rate climbs 20-25 points, and depreciation recovery climbs 30+ points. None of those gains come from AI "negotiating" — they come from AI making the documentation tight enough that the work argues for itself.
Want to model your own numbers?
The ROI calculator includes claim-side inputs: monthly claim volume, average ticket, current supplement approval rate. Drag the sliders to see what claim-documentation AI does for your specific shop.
Common objections
"My adjusters will know I'm using AI."
Some will. Most won't care. What they care about is whether your documentation is good. The contractors who already submit clean, Xactimate-friendly photo packages with line-item-level justification get faster approvals. AI just lets you produce that quality at scale, on every claim, instead of only on your top 30%.
"My public adjuster won't like it."
Mine like it. So do every PA we've worked with. AI handles documentation and prep — exactly the part PAs hate doing manually. Their negotiation work gets easier and faster because the evidence is pre-organized. Most PAs we've talked to want their contractor partners on these systems.
"What if AI makes a mistake on a line item?"
Same answer as for any other AI workflow: human review before send. Every supplement letter the AI drafts gets a 60-second human eyeball before it goes to the carrier. Misclassifications happen on maybe 4-7% of line items in our deployments, and they're caught at the review stage. The cost of catching them is far less than the time saved on the 93%+ that are correct.
"Storm restoration is a relationship business — AI feels cold."
The relationship parts (homeowner trust, the adjuster meeting, the negotiation) are still 100% human. AI doesn't touch those. It does the back-office documentation and admin work — the part nobody enjoys and most contractors do badly under volume pressure. Customers don't notice. Adjusters don't care. Your team gets their evenings back.
How to actually deploy this
Week 1: Setup & integration
- Connect CRM (JobNimbus / AccuLynx / similar — we've documented the JobNimbus version on the JobNimbus integration page)
- Connect photo source (CompanyCam, drone, mobile)
- Connect measurement source (Hover, EagleView, manual)
- Configure your supplement letter templates
- Configure your Xactimate-friendly scope language
Week 2: First 5 claims in shadow mode
- AI processes incoming claims, generates documentation packets
- Your team reviews each one, edits, sends
- Voice and language tuning — making sure scope language matches your shop's standard
Week 3: First batch live
- Production deployment — all incoming claims flow through automation
- Human review is now a 60-90 second checkpoint, not a 4-hour build
- Supplement workflow goes live
Week 4: Tracking + tuning
- Recoverable depreciation tracker active for all completed jobs
- Re-inspection scheduling and prep automated
- Daily summary email going to ops/owner
FAQ
Is this legal?
Yes — within limits. AI handles documentation, scope, and supplement drafting. It doesn't act as a public adjuster. Stay on your side of the line (covered above in the where-AI-is-allowed section) and you're fine. When in doubt, your public adjuster handles negotiation; you handle the roof.
Does it work with Xactimate?
Yes. AI produces output formatted to Xactimate line items — same item codes, units, and scope language. AI can also analyze a denied or under-paid Xactimate estimate and generate a supplement letter draft. AI doesn't replace Xactimate; it produces inputs Xactimate (and the adjuster reviewing it) can process more efficiently.
What about Hover and EagleView integration?
Hover, EagleView, and similar measurement reports are excellent inputs to AI claim documentation workflows. AI ingests measurement data, combines it with photo evidence, generates a damage map with line-item descriptions, and produces a structured packet ready for adjuster review or supplement drafting.
Can AI talk to adjusters?
No, and this is a hard line. AI can prepare you for adjuster conversations and document everything that's said — but it shouldn't be the actual voice on the phone or in correspondence on behalf of a homeowner. Your team or your public adjuster handles the conversation; AI handles the prep and documentation.
What does setup cost?
Implementation through the AI Clarity Sprint runs $2,500 flat for full setup — documentation workflows, supplement-letter templates, CRM integration. Tooling runs $200-500/month for a storm-restoration shop processing 30-100 claims/month. Per-claim variable cost is $3-12. Payback is typically inside the first 30 days from documentation time savings alone.
How does this play with my CRM?
JobNimbus and AccuLynx are the most common — both have native integration. The full claim journey runs inside your existing CRM: claims tracked, documents auto-saved, photo packages generated, supplement letters drafted in your branded format. Full JobNimbus workflow is on the JobNimbus integration page.
What if I don't have a public adjuster relationship?
This system works either way — and it makes contractors who handle their own documentation (without a PA) significantly stronger at the table. But for any state with strict PA regulations (Texas, Florida, California, etc.), if you're charging fees for claim representation, you need a licensed PA in the loop. AI strengthens that relationship; it doesn't replace it.
What to do with this
Three options:
- Run your numbers. Open the cost calculator, plug in your monthly claim volume and average claim RCV, and see what the documentation + supplement automation does for your shop.
- Get the playbook. The full Roofer's AI Playbook includes the storm-system architecture and the templates we use on every storm-restoration deployment.
- Book a 30-min audit. We'll review your current claim workflow, identify the documentation steps that automate cleanly, and quantify the supplement-recovery upside for your specific volume. Schedule here.
The reality: storm restoration in 2026 is increasingly a documentation race. Carriers are tightening. Adjusters are stricter. The contractors who consistently capture full RCV are the ones with documentation tight enough that the work speaks for itself. AI is the cheapest way to get there.
Book a 30-min audit →