What changed
FACT (from source text): a $3,569,910,110.67 Department of Homeland Security assistance award was made to a Louisiana homeland security & emergency agency to 'PROVIDE REIMBURSEMENT TO STATE, LOCAL, TRIBAL, AND TERRITORIAL GOVERNMENT ENTITIES AND CERTAIN PRIVATE NON-PROFIT ORGANIZATIONS FOR EMERGENCY PROTECTIVE MEASURES TAKEN DURING THE PANDEMIC,' explicitly enumerating emergency medical care, medical sheltering, and administration and distribution of vaccines. INFERENCE: this is FEMA Public Assistance Category B, delivered through the FEMA Grants Portal with GOHSEP as recipient/pass-through and parishes, tribes and PNPs as subrecipients. CAUTION (unverified, flagged): the award identifier embeds '4611DR-LA', which does not obviously correspond to the COVID-19 major disaster declaration for Louisiana; the disaster number and the pandemic description in the text are in tension and must be reconciled before any build. Nothing in the source text states a deadline, a subrecipient count, or that any claims remain unfiled.
Why now
The honest answer is that 'now' is weak, and this is the central problem with the idea. HYPOTHESIS, not fact: the pandemic Category B claim window is largely retrospective — obligations, project worksheets and most closeouts for COVID-era emergency protective measures were substantially worked through in the years after the declarations. What plausibly remains is a tail: requests for information, second appeals, deobligation disputes, and OIG / 2 CFR 200 Subpart F Single Audit findings. A tail is a real market but it is a shrinking one with no forward deadline, which is the opposite of the founder's proven FMCSA ELDT shape, where a live mandate produced a perpetual stream of new filers. The source text states no deadline ('DEADLINE: none stated in text'), and I will not manufacture one.
Converging signals
Only one signal in the input is actually about this opportunity: the DHS/FEMA award of $3,569,910,110.67 to the Louisiana homeland security and emergency agency. The other thirty 'FUNDED MANDATE' items retrieved as demand evidence are Medicaid Title XIX entitlement grants to state Medicaid agencies (PA DHS, NYS DOH, CA DHCS, LA Department of Health, etc.) and unrelated DHS disaster-damage repair grants to the Virgin Islands, Puerto Rico and New York. Those are semantic-similarity artifacts — 'Department of Health and Human Services' and 'grant to local government' embed near 'healthcare' and 'FEMA award' — and they carry zero information about whether a PNP hospital will pay for Category B substantiation software. Scoring convergence off thirty near-duplicate Medicaid capitation awards would be exactly the fabrication failure the system's own lesson (confidence 0.895) warns about. Real convergence here is therefore one signal, not three.
Customer pain
INFERENCE, supported by the enumerated cost categories in the award text rather than by any complaint evidence in the input: a PNP hospital that stood up alternate care sites, paid overtime for surge staffing, and ran vaccine clinics must, to sustain reimbursement, produce per-employee payroll and overtime records tied to specific eligible work, procurement files demonstrating 2 CFR 200.317-327 compliance for emergency contracts, sheltering and vaccination logs, and a per-line-item narrative connecting each cost to an eligible emergency protective measure. PNPs are the weakest filer class here: unlike a parish or a state agency they have no emergency-management staff, no grants office, and no institutional memory of FEMA documentation standards. That much is a reasonable inference. What the input does NOT contain is a single complaint, forum post, job posting, or consultant advertisement proving a PNP is currently in pain and looking to spend. The demand_evidence array contains no PAIN item and no HIRING/SPEND item relevant to this idea.
Who pays
Candidate buyers, in descending order of reachability: (1) small and mid-size disaster-recovery consultancies who already hold PNP engagements and would use the tool to lower their own labor cost; (2) PNP hospital finance / compliance staff facing an active RFI or audit finding; (3) parish and tribal subrecipients. Buyer (2) is the nominal target and is the weakest commercially — a hospital finance office buys software through a committee, on an annual cycle, and only when there is an active, expensive threat. Buyer (1) is more realistic but is a channel of a few dozen firms nationally, and they build internally or use Deltek/Excel.
Solved today
Specialist disaster-recovery consultancies — Hagerty Consulting, Tidal Basin, Witt O'Brien's, IEM, AC Disaster Consulting — plus regional CPA firms doing Single Audit work. FEMA's own Grants Portal is the system of record for project worksheets and document upload. Juvare's Crisis Track and similar products already do FEMA-oriented damage assessment and cost capture. HYPOTHESIS: most PNPs did this in spreadsheets with consultant help, and the consultant's fee was itself charged to the grant.
Why current solutions are bad
Spreadsheet-and-shared-drive substantiation loses the audit trail between a dollar and its eligibility justification, which is precisely what OIG findings turn on. That is a genuine product-shaped defect. But 'bad' is not the same as 'the buyer will change,' and this is where the idea starts to fall apart — see kill arguments.
Proposed product
A structured workspace: ingest payroll exports, timekeeping data, invoices and contracts; map each cost line to a Category B eligible-work taxonomy; force a narrative field per line with a checklist derived from FEMA PA Program and Policy Guide criteria and 2 CFR 200 procurement standards; flag the recurring OIG deobligation triggers (costs outside the incident period, non-competitively procured contracts without a valid emergency justification, duplication of benefit against Provider Relief Fund / HHS / insurance, undocumented overtime differentials); export an assembled, indexed package ready for Grants Portal upload and for an auditor's document request. Critically, this is a preparation tool, not a submission robot — which is the difference between it and the founder's FMCSA ELDT product.
MVP version
Duplication-of-benefit and eligibility triage on a single applicant's cost ledger: upload a payroll/GL extract, get a line-by-line risk score with the specific FEMA/2 CFR citation that would sustain or deobligate each line, plus a generated narrative draft per line. Ship it as a paid one-time assessment before building any workspace.
30-day build
Do NOT build. Spend the month falsifying the premise. Reconcile the 4611DR-LA identifier against FEMA's declaration and PA obligation data. Pull FEMA's public Public Assistance Funded Projects Details dataset and count how many Louisiana PNP applicants have open (unobligated or appealed) Category B projects, versus closed. Read the last three years of DHS OIG audit reports on COVID Category B to extract the actual deobligation reasons. Call six disaster-recovery consultants and ask what they bill, who pays, and whether that fee is reimbursed. If the count of open PNP projects is in the low dozens, kill.
60-day build
Only if the 30-day count survives: hand-deliver paid assessments. Take three PNP files, do the substantiation manually with AI assistance, charge $5,000-$15,000 each, and instrument which steps consumed the hours. The product spec falls out of that. This is consulting-first, deliberately, because the software thesis is unproven.
90-day revenue plan
Convert the manual work into the triage tool and sell it to the consultancies who do this at volume, not to the PNPs. Revenue by day 90 is plausible only from services. Software revenue by day 180 is a hypothesis, not a plan. I am not going to dress a services engagement up as a micro-SaaS ramp.
Distribution path
Weak, and this is a second-order kill argument. There is no list of 'PNPs with open Category B claims' that a stranger can cold-email into a purchase. FEMA's PA applicant data gives names, not decision-makers. The reachable channel is the two dozen disaster-recovery consultancies and the regional CPA firms doing Single Audits — a relationship sale to a small, incumbent-friendly set, which is explicitly the sales motion the founder avoids. Demonstrated-value distribution (publish the OIG deobligation-reason analysis as a free artifact and let filers find it) is the one honest path, and it is slow.
Pricing hypothesis
$5,000-$15,000 per applicant engagement for the assessment; $500-$1,500/month per active project for the workspace. The proposed '% of reimbursement sustained on OIG audit' contingency model in the input should be dropped: contingency fees tied to obtaining or retaining federal grant funds sit in an unallowable-cost gray zone under 2 CFR 200, meaning the fee could not be charged to the grant — which removes the one reason the buyer would agree to it. HYPOTHESIS, worth legal review, but the risk is real enough to price around.
Technical difficulty
Low to moderate. Document ingestion, an eligibility rules layer, LLM-drafted narratives with citation. There is no portal API to reverse-engineer, which cuts both ways: it is easy to build and equally easy for anyone else to build. The FEMA Grants Portal exposes no public submission API, so the founder cannot replicate the FMCSA ELDT wedge of being the pipe into the government system. Without that pipe, the product is a smart document assembler, and the moat is the eligibility rules content, not the integration.
Legal / regulatory risk
Moderate. Preparing federal grant substantiation that a subrecipient certifies to carries False Claims Act exposure if the tool's narratives overstate eligibility; the founder would want contractual clarity that the applicant certifies, not the vendor. Contingency pricing on federal funds is likely unallowable. Automating interaction with the Grants Portal beyond manual upload would touch federal system rules of behavior. None of these are fatal; all are real.
Platform dependency
None in the deplatforming sense — FEMA cannot ban the tool, and per the scoring rules I am not flagging platform_policy_risk. But there is a policy-obsolescence dependency: FEMA could issue simplified closeout guidance, or the COVID PA tail could be administratively wound down, and the market disappears without any competitor doing anything.
Founder fit
Genuinely strong on capability, weak on shape. His fire-service and emergency-operations background means he understands incident periods, mutual aid, overtime differentials and what an alternate care site actually is — that is rare and it is the reason to take this seriously at all. Public-records fluency maps onto FEMA PA disclosure data. But the shape he has proven — a live mandate forces a perpetual class of filers to submit into a portal, and he owns the pipe and charges per submission — is absent here. This is retrospective, has no portal API, no deadline, no new filers, and an incumbent consultant class whose fee the buyer does not pay out of pocket. Fit on domain: 8. Fit on business shape: 4.
Breakout potential
The interesting asset is not the Louisiana COVID tail — it is the eligibility rules engine. Every declared disaster produces Category A-B claims forever, roughly 50-60 major declarations a year, each with a fresh cohort of unsophisticated PNP and small-jurisdiction applicants under real deadlines. That is a live, recurring, deadline-bearing market with genuinely forced filers. If the founder wants this space, that is the product, and the COVID audit tail is at best a way to build the rules content and get three reference customers. Note that Juvare's Crisis Track and the large consultancies already occupy that ground, so 'live disaster' is the better market AND the more crowded one.
Final recommendation
DO NOT BUILD as specified. The public-money shape is present on the surface and absent underneath: real money, real filer class, real paperwork — but no forward deadline, no submission API to own, and an incumbent whose fee the buyer does not pay. The reimbursable-consultant-fee argument alone is close to fatal, and I could not construct a rebuttal to it. Score it as a strong domain match wrapped around a weak business. The one thing worth salvaging is the founder's emergency-operations fluency applied to LIVE disaster declarations rather than the COVID tail, where filers are genuinely forced and deadlines genuinely bind — and even there he would be entering an occupied market, so it deserves its own brief and its own kill attempt rather than a rescue clause bolted onto this one. Recommend REVISIT LATER, gated on the 30-day falsification work.
Next action
Before writing a line of code: download FEMA's public Public Assistance Funded Projects Details dataset, filter to Louisiana COVID-era declarations and PNP applicants, and count projects that are still open, appealed, or unobligated. Simultaneously confirm from FEMA policy whether subrecipient management/administrative costs — including consultant fees — are reimbursable under this award. If open PNP projects number in the low dozens, or if consultant fees are grant-reimbursable (which I expect), kill the idea and redirect the same domain knowledge at live disaster declarations.