What changed
FACT (from source text): DHS awarded $3,171,904,245.51 to a Louisiana homeland security & emergency agency for repair/replacement of disaster-damaged facilities under DR-4559-LA (Hurricane Laura), award ID ASST_NON_4559DRLAP00000001_070. Comparable DHS awards appear for the Virgin Islands ($21,985,858,464.89, DR-4340), Puerto Rico ($35,301,159,434.96, DR-4339) and New York State DHSES ($17,365,135,822.49, DR-4480). INFERENCE: these are state-level grantee obligations that pass through to hundreds of local subrecipients, each of whom must individually document their claim. The dollar figures as returned are cumulative/aggregate award ceilings and should not be read as a single obligation to one project.
Why now
INFERENCE, not established by the source text: FEMA PA obligations from the 2020-2022 disaster years are in the project-formulation and closeout phase, which is exactly when insurance reconciliation becomes the binding constraint on obligation. The source text states no deadline. The 'why now' is therefore weak-to-moderate: this is a standing, structural obligation created by the Stafford Act rather than a newly opened window. Treat any claim of urgency as hypothesis.
Converging signals
Three things meet at one point: (1) a very large appropriated pool of federal disaster money already obligated to state grantees (FACT, four USAspending awards cited); (2) a defined class of forced filers — Louisiana parishes, municipalities, school boards and private non-profits filing Requests for Public Assistance and Project Worksheets with GOHSEP (asserted in the convergence description; the underlying award record itself only names the state recipient, so the filer class is INFERENCE from how FEMA PA is structured); (3) a specific, mechanical, error-prone sub-task within that filing — reducing the claim by actual and anticipated insurance proceeds — that is currently done by hand in spreadsheets.
Customer pain
HYPOTHESIS (no complaint threads, job postings or PRA respondent counts were supplied in the input; demand_evidence contains only award records). The asserted pain: a subrecipient with a damaged school, courthouse or lift station must reconcile a commercial property policy — often with blanket limits, sublimits, deductibles per-occurrence vs per-building, and a partial settlement that does not map to FEMA's damage inventory — against FEMA's site-by-site cost estimate. Get it wrong and the money is deobligated years later at closeout or in an OIG audit, and the local government repays. That repayment risk is the pain, and it is real in kind even though the input does not evidence its intensity.
Who pays
Three candidate buyers, in descending order of reachability. (1) The PA grant-management consultancies — Hagerty, Tidal Basin, Witt O'Brien's, IEM, CDR Maguire, Ernst & Young's public-sector practice — who are retained by subrecipients and bill hourly or as a percentage of award. They buy tools that cut labor hours on a fixed-fee engagement. (2) Public-entity insurance pools and their brokers, who hold the policy data and get blamed when the FEMA offset is wrong. (3) The subrecipients themselves — the least reachable, because they buy through municipal procurement on annual budget cycles.
Solved today
INFERENCE: today this is a consultant with Excel, a PDF of the policy, the adjuster's Statement of Loss, and FEMA's Damage Inventory exported from Grants Portal. The reconciliation is assembled by hand into a workpaper that FEMA's insurance specialist reviews. FEMA itself employs insurance specialists who perform an independent review. Consultants bill for this work; that billing is the existing spend, and it is inferred, not evidenced by the input.
Why current solutions are bad
HYPOTHESIS: manual mapping of settlement line items to FEMA damage-inventory line items is slow, non-reproducible, and does not survive audit three years later when staff have turned over. There is no artifact that shows the derivation. Nothing in the provided sources establishes this; it is asserted from domain structure.
Proposed product
A document-in, workpaper-out reconciliation tool. Upload the declarations page, the policy form, the adjuster's Statement of Loss, any partial settlement letters, and the FEMA Damage Inventory export. The tool extracts limits, sublimits, deductibles and paid amounts; maps settlement line items to damage-inventory sites; applies the deductible allocation and anticipated-proceeds logic; and emits (a) a per-site eligible net claim, (b) an auditable derivation showing every offset and its source document and page, and (c) an exception list of the line items it could not confidently map, flagged for human review. It never files anything and never asserts an answer it cannot cite to a page.
MVP version
Ingest 20-40 real historical PA insurance workpapers (obtainable via public-records requests to GOHSEP and to Louisiana parishes — the founder's public-records strength applies directly). Build extraction for declarations pages and Statements of Loss. Build the mapping and offset engine. Emit the workpaper as XLSX plus a PDF derivation appendix. Ship it as a desktop-or-web tool a consultant runs per project. No portal integration in the MVP — FEMA Grants Portal exposes no public submission API, and building against it is not required for the wedge.
30-day build
Do not write the mapping engine yet. File public-records requests with GOHSEP and three Louisiana parishes for closed PA insurance workpapers and any deobligation/OIG findings tied to insurance. Interview eight people: two GOHSEP applicant-services staff, three consultants at the firms named above, two public-entity pool risk managers, one FEMA insurance specialist if reachable. The single question to answer: does the consultant's fixed-fee engagement actually make this labor a cost they want to remove, or is it billable time they want to keep? If it is billable time they want to keep, the primary channel is dead and the product must be sold to the pools or the subrecipients instead. Kill or pivot on that answer.
60-day build
Assuming the channel survives: build extraction plus the offset engine against the corpus. Recruit two consultancies as design partners on a paid pilot (they pay, or it is not validation). Instrument accuracy against the historical workpapers — the metric is percent of settlement line items correctly mapped without human correction, and the product is not sellable below roughly 80% with a clean exception list.
90-day revenue plan
Per-project pricing to consultancies, invoiced against their engagement. First revenue is realistically a pilot invoice, not a self-serve signup. A 90-day revenue target here is optimistic; 120-180 days is the honest number, which the founder's stated runway tolerates.
Distribution path
Direct outreach to the twenty-odd firms that hold PA grant-management contracts — a countable, named, reachable list, not a market to be advertised into. Secondary: GOHSEP applicant briefings, the Louisiana Municipal Association, and IAEM/NEMA conferences. This is a small-list, high-touch sale. It is not enterprise procurement, but it is relationship-adjacent, and the founder's stated preference for selling through demonstrated value works only if he can put a correct reconciliation of a real, public, historical claim in front of a consultant on the first call.
Pricing hypothesis
$1,500-$4,000 per project reconciliation, or an annual seat for a consultancy running many. Explicitly not a percentage of award: taking a percentage of federal grant money is a contingency-fee arrangement that federal cost principles (2 CFR 200) treat as an unallowable cost, so the subrecipient cannot charge it back to the grant. The convergence description's proposed '% of award protected' monetization should be discarded.
Technical difficulty
Moderate and concentrated in one place. The extraction of policy limits and settlement line items from heterogeneous PDFs is the whole product; the arithmetic is trivial. Insurance policy forms are not standardized across carriers, and blanket limits with sublimits are genuinely hard to reduce to per-site allocations. Expect the exception rate, not the happy path, to determine whether anyone pays.
Legal / regulatory risk
Real and under-appreciated. Computing an eligible net claim that a subrecipient relies on, which is later deobligated, invites a claim against the vendor. Mitigations: position as a workpaper-preparation tool that shows its derivation, never as an opinion; require human sign-off on every output; carry E&O insurance; keep an exception list that forces review. Adjusting insurance claims for a fee is a licensed activity in Louisiana and most states — the product must stay clearly on the grant-documentation side of that line and not advise on the insurance claim itself. This is a case where the founder himself could be pulled toward needing a license, which is exactly the condition the founder profile says to avoid.
Platform dependency
None meaningful. There is no platform owner who can deplatform a tool that produces a document. FEMA Grants Portal has no API to depend on, which is a limitation rather than a dependency.
Founder fit
Strong but not maximal, and weaker than the surface pattern-match suggests. It matches the ELDT precedent in shape — a federal program, a defined filer class, paperwork, a per-transaction fee — but differs in the load-bearing respect: ELDT was a per-upload transaction against a portal with a submission API-equivalent and an individual, self-serve buyer. This has no submission step, no self-serve buyer, and a sale into a small list of consulting firms. The transferable strengths are the public-records sourcing of a training corpus, the operational credibility with parish and emergency-management staff from the fire-service background, and comfort with regulatory text.
Breakout potential
Good if the wedge holds. The same engine generalizes to every state's PA program and to HUD CDBG-DR, where duplication-of-benefits reconciliation is a statutory requirement under the Stafford Act §312 and has produced repeated HUD OIG findings. Fifty near-identical state markets and a second federal program is real expansion, and it is expansion by replication rather than by network effect, which suits the founder.
Final recommendation
CONDITIONAL — do the 30-day channel-validation work before writing any product code, and be genuinely willing to kill it. The regulatory shape is right and the founder's public-records and emergency-management background give him an unusual ability to build the training corpus and get the first meetings. But the demand in this input is inferred from award records, not evidenced, and the entire business rests on one unverified assumption: that PA consultancies are on fixed-fee engagements and therefore want their own labor removed. That assumption is checkable in three phone calls and one look at a publicly-procured PA grant-management contract. Check it first. If consultancies bill hourly, this reduces to selling municipal software through procurement, and should be dropped.
Next action
Pull two or three publicly-posted GOHSEP or Louisiana parish PA grant-management contracts and read the compensation clause. Fixed-fee or not-to-exceed means the channel is alive; hourly or cost-plus means it is not. Do this before any other work — it costs a day and it determines whether the idea exists.