What changed
FACT (source: expressnews.com): a news report states USAA closed 51% of home insurance claims in 2025 without any payment, surfacing a large, fresh pool of denied homeowners. INFERENCE: cheap open-weight LLM inference (DeepSeek/GLM-class, ~29% of production volume per the Vercel index) now makes parsing denial letters and drafting rebuttals near-zero marginal cost for a solo builder.
Why now
Fresh, quantified reporting on mass no-pay closures creates a concentrated, money-motivated, angry audience RIGHT NOW; simultaneously per-token cost has collapsed so the drafting can be automated cheaply. Both are cited FACTS.
Converging signals
A consumer-pain complaint signal (mass no-pay denials) meets a cheap-capability signal (open-weight LLM drafting). This is a pain × capability quick-win convergence, NOT a public-money/forced-buyer mandate — score it on the discretionary rubric.
Customer pain
HYPOTHESIS grounded in the cited denial-rate fact: a homeowner holding a denial or lowball letter for thousands of dollars of damage, who does not know the appeal process, the statutory timelines, or how to rebut the carrier's specific cited reason. The pain is acute and dollar-quantified, but the volume/intensity of INDIVIDUAL complaints is inferred, not directly evidenced (demand_evidence array is empty).
Who pays
Homeowners with a recently denied or underpaid property-damage claim. Adjacent, higher-value buyers: public adjusters, restoration/roofing contractors, and small policyholder-advocacy shops who could white-label the packet generator per-claim.
Solved today
Free ChatGPT prompts; public adjusters (typically 10-20% contingency of the recovered amount); attorneys; DIY letters; or giving up. State DOI complaint portals offer a free appeal path.
Why current solutions are bad
Public adjusters take a large percentage and won't touch small claims; attorneys are disproportionate to a few-thousand-dollar dispute; a generic ChatGPT letter does not cite the specific policy clause, the carrier's stated reason, or the state's appeal statute/deadline — which is exactly the KILL TEST this idea must beat.
Proposed product
A narrow web tool: upload denial letter + policy + damage photos → LLM extracts the cited denial reason, maps it to the matching policy language, and generates an appeal letter that cites the state's insurance appeal statute/DOI complaint route, the relevant policy clause, and an evidence checklist tuned to that carrier's stated reason. Defensibility lives in the carrier-and-reason-specific templates, state statute/deadline library, and DOI escalation instructions — the structured legal/procedural scaffolding a raw prompt lacks.
MVP version
Single-page upload + LLM extraction + templated letter for the top 3-4 denial reasons (wear-and-tear/maintenance exclusion, insufficient documentation, cause-of-loss dispute, late notice) for one or two high-volume carriers in 2-3 states. Stripe checkout at $49. Buildable in days-to-2-weeks on cheap inference.
30-day build
Ship MVP for 2 states + top carriers; hand-curate statute/deadline/DOI-complaint data for those states; validate that generated packets are materially better than a raw ChatGPT letter (blind comparison); seed via denied-claim Reddit/Facebook groups and public-adjuster forums.
60-day build
Expand carrier + denial-reason coverage and to ~10 states; add the $99 tier (follow-up rebuttal + DOI complaint draft); pursue contractor/restoration and public-adjuster white-label deals for recurring volume.
90-day revenue plan
Revenue from direct $49/$99 packets plus 1-3 white-label/referral partners feeding claims; realistic path to first revenue within days-to-weeks given zero procurement, but honest scale depends on cheap distribution to a hard-to-reach, one-time-need audience.
Distribution path
Content/SEO on '[carrier] denied my claim what to do', denied-homeowner Reddit/Facebook groups, DOI-complaint how-to content, and B2B2C via contractors/public adjusters who see denials daily.
Pricing hypothesis
$49 per appeal packet; $99 with a follow-up rebuttal + DOI complaint draft; optional white-label per-seat/per-claim for pros.
Technical difficulty
Low. Document parsing + templated generation on cheap open-weight inference. The hard part is not code — it is curating accurate, current state statutes, deadlines, and carrier-reason mappings, and not drifting into unauthorized practice of law.
Legal / regulatory risk
Real and the primary risk. Drafting legal-ish letters and citing statutes flirts with unauthorized practice of law (UPL) in some states; must be positioned as a self-help document-preparation tool with clear disclaimers, no representation, no legal advice. Bad statute citations could harm a user's claim — accuracy liability. Not licensure-BLOCKING if scoped as self-help doc prep, but must be handled carefully.
Platform dependency
None material — no app-store or government-portal gatekeeper. Depends on an LLM inference provider, but that is swappable across open-weight vendors.
Founder fit
Moderate. Plays to complaint-mining, AI workflows, fast prototyping, and public-records/statute assembly — all founder strengths. But it is NOT the founder's highest-fit shape: there is no government portal to submit to, no forced-buyer class, and no per-filing mandate. It is a discretionary consumer painkiller with a one-time, hard-to-retain buyer.
Breakout potential
Moderate. Replicable across 50 states and every carrier, and extensible to auto/health/denied-benefit appeals (a large adjacent family). Ceiling is capped by one-time buyers, churn, and the UPL/accuracy overhang; the durable version is the B2B2C white-label to pros with recurring volume.
Final recommendation
BUILD-AND-TEST, low-cost validation first. The pain is real and dollar-quantified and the build is genuinely small and cheap, so it clears the quick-win bar for a fast MVP. But treat it as a probe, not a conviction bet: before scaling, run the blind ChatGPT-vs-ClaimAppeal comparison to prove the wrapper is defensibly better, and validate that denied homeowners can be reached cheaply. It is NOT the founder's top-fit public-money/forced-filer shape — pursue only alongside, not instead of, mandate-shaped opportunities.
Next action
Hand-build the packet generator for ONE carrier (USAA) × the top 2 denial reasons in ONE state, then blind-test its output against a raw ChatGPT letter with 3-5 denied homeowners recruited from a denied-claim Reddit/Facebook group; charge the first $49 to confirm willingness to pay before expanding coverage.