What changed
FACT (Federal Register 2026-08494, final rule published 2026-05-01): the CFPB finalized revisions to Regulation B subpart B implementing Dodd-Frank Section 1071, amending institution/transaction coverage, the small business definition, required data points, and the compliance date. Every lender that remains covered must rework its 1071 data collection and reporting before the new compliance date. Concurrently (FACT, cited sources): PP-OCRv6 provides sub-35M-parameter self-hostable OCR, and GPT-5.6-class models lower per-document extraction cost — making a one-developer extract-validate-report pipeline economically feasible.
Why now
The final rule (May 2026) ends a multi-year period of uncertainty (2023 rule → litigation → 2024/2025 interim rules → 2025 compliance-date extension → 2025 proposal → 2026 final). Lenders who deferred spending because 'the rule keeps changing' now face a fixed target and must build or buy. HYPOTHESIS: a post-finalization buying window of roughly 6-18 months exists before the first compliance tier hits. CAVEAT (inference from the cited docket history): dates have been extended multiple times and litigation risk persists, which can drain urgency again — this is the single biggest timing risk.
Converging signals
(1) Regulatory forcing function: finalized 1071 revisions with a compliance date [FACT, federalregister.gov 2026-08494]. (2) Cheap self-hosted OCR: PP-OCRv6, 1.5M-34.5M params, 50 languages [FACT, huggingface.co]. (3) Cheaper frontier reasoning for structured extraction/validation [FACT as to the release claim, openai.com/index/gpt-5-6; unit-economics benefit is inference]. Signals 2+3 collapse the build cost of what was previously an enterprise-vendor-only product; signal 1 supplies the forced buyer.
Customer pain
HYPOTHESIS grounded in the rule's structure: covered lenders must collect ~dozens of data points per covered application (demographics, pricing, action taken, etc.), many living in unstructured loan files, PDFs, and LOS free-text fields. Commercial/small-business lending shops — unlike mortgage shops with 40 years of HMDA muscle memory — often have no data-collection discipline for this. The 2026 revisions changed which data points and which institutions are in scope, so even lenders who started building against the 2023 rule must redo mapping and validation. NOTE HONESTLY: the demand_evidence array contains only FORCED BUYER mandate documents — zero PAIN complaints and zero HIRING/SPEND postings were retrieved — so willingness-to-pay at the small-lender tier is inferred from the mandate, not observed.
Who pays
Covered small-business lenders at the bottom of the market: community banks under ~$1B, credit unions, CDFIs, farm credit institutions, equipment finance and independent commercial finance companies — institutions above the (revised) origination threshold but too small for a compliance department or a six-figure Wolters Kluwer/Ncontracts contract. Buyer persona: the compliance officer or CFO who is personally responsible for the filing. IMPORTANT CAVEAT (inference from the 2025 proposal's stated intent to 'reconsider coverage'): the 2026 final rule likely narrowed coverage versus 2023, meaning the smallest lenders — this product's natural buyers — may have been exempted. Sizing the surviving covered population against the final thresholds is the first validation task, before any code.
Solved today
HYPOTHESIS (well-supported by public vendor marketing, not in provided sources): incumbent compliance-software vendors (Ncontracts/TRUPOINT, Wolters Kluwer, Asurity RiskExec, ACES) sell 1071 modules to their existing HMDA/CRA customer base; larger LOS platforms (nCino, Abrigo) are adding native 1071 collection; consultants (Wipfli, crowe-tier firms) sell readiness assessments; the smallest lenders plan to do it in Excel plus manual file review.
Why current solutions are bad
Incumbent tools assume the data is already structured in an LOS and priced/packaged for institutions that already buy their suite — they do compliance workflow, not extraction from messy loan files. Excel-plus-intern collapses under the validation-edit rules and the demographic-data firewall requirements. The revised rule creates a re-mapping problem even for lenders who bought tooling against the 2023 version. The gap: nobody serves 'we have 300 covered applications a year sitting in PDFs and a part-time compliance person' with a low-touch, per-season-priced product.
Proposed product
A hosted (or on-prem-deployable, using the self-hostable OCR for data-security-sensitive buyers) pipeline: ingest loan files/LOS exports → PP-OCRv6 + LLM extraction of the required 1071 data points → deterministic validation against the final rule's data-point spec and edit rules → exception queue for human review → produce a submission-ready file for the CFPB filing platform, plus an audit trail documenting how each field was derived. Wedge features: the revised-rule delta (2023→2026 field mapping), firewall-compliant handling of demographic data, and a board-ready readiness report.
MVP version
A '1071 File Readiness Scan': lender uploads 20 sample loan files, tool returns a scored gap report — which required data points are capturable, which are missing, per-file extraction with confidence flags — plus the validated CSV. This is sellable as a $1-2.5k fixed-price assessment before the full pipeline exists, and every scan trains the extraction prompts. Build: FastAPI + PP-OCRv6 + LLM extraction with a hand-built validation rule set transcribed from the final rule's data-point appendix. 4-8 weeks solo with AI assistance.
30-day build
(1) Read the final rule's coverage thresholds and compliance-date tiers; confirm the surviving covered-lender population and the real deadline calendar — kill or proceed on this. (2) Interview 10 compliance officers at CDFIs/community banks/equipment finance shops (state banker associations, NACHA/CBA compliance forums, LinkedIn) asking one question: 'what is your 1071 plan and budget?' (3) Transcribe the data-point/validation spec into a machine-readable rule set. (4) Prototype extraction on synthetic loan files.
60-day build
Ship the Readiness Scan MVP. Sell 5 paid scans at $1-2.5k via direct outreach and one webinar co-hosted with a state banking association or CDFI network (these associations actively need 1071 content for members — HYPOTHESIS but low-cost to test). Use scan results to spec the full pipeline and lock pricing.
90-day revenue plan
Convert scan customers to annual pipeline subscriptions ($4-9k/yr per institution, sized well under incumbent suites) or per-application pricing (~$3-8/covered application) mirroring the founder's proven ELDT per-filing model. Target: $10-25k booked by day 120-180. Revenue inside 180 days is plausible only if the deadline calendar confirmed in the first 30 days shows lenders must be collecting data within ~12 months — otherwise urgency evaporates and this becomes a 2027 revenue play.
Distribution path
Direct, demonstration-led (matches founder's selling style): free '1071 delta cheat sheet' (2023 vs 2026 final rule field-by-field) as lead magnet; state banking association newsletters and webinars; CDFI coalitions (OFN); LinkedIn outreach to compliance officers; possibly white-label through the small compliance-consulting firms that lack software. No ad spend, no marketplace.
Pricing hypothesis
Readiness Scan $1-2.5k one-time → annual subscription $4-9k/yr or per-application fee. Anchor against incumbent suite pricing (tens of thousands) and against the cost of a consultant (hundreds/hr). Per-filing pricing is the founder's proven motion.
Technical difficulty
Moderate. OCR+LLM extraction from heterogeneous loan files is genuinely hard to make reliable (expect a human-in-the-loop exception queue, not full automation — promising full automation would be dishonest and dangerous in a fair-lending context). Validation rules are deterministic drudgework. The CFPB filing platform's submission spec must be tracked. Solo-buildable with the founder's AI-workflow skills: yes.
Legal / regulatory risk
Real but manageable: the tool touches ECOA/fair-lending data, including demographic data subject to the firewall provision — errors have regulatory consequences for the CUSTOMER, so contracts must position this as a data-preparation tool with lender-reviewed output, not legal advice or a compliance guarantee. Borrower loan files are GLBA-sensitive nonpublic personal information: expect vendor due-diligence questionnaires, need for solid security posture, eventually SOC 2 (a fundable expense given runway, but a real drag on solo sales velocity). Also: ongoing litigation/rulemaking history means the rule itself could shift again — that risk is symmetric (it also resets incumbents).
Platform dependency
Low. No app store, no third-party API gatekeeper. Dependency is on the CFPB's filing spec and platform remaining stable — the same class of dependency the founder already manages with the FMCSA TPR.
Founder fit
HIGH on shape, with one honest deduction. This is exactly the proven pattern (lesson, confidence 0.80, applies): a federal mandate compels a class of filers to submit data to a government system; founder builds the submission/automation layer and charges per filing — direct rhyme with his shipped FMCSA ELDT product. Systems thinking, document automation, public-records fluency all apply. THE DEDUCTION: the buyer is a regulated financial institution, not a trucking school — vendor due diligence and institutional conservatism make this a slower, higher-trust sale than ELDT, which grates against his 'no multi-year trust-building' preference. Targeting CDFIs and non-bank commercial finance companies (lighter procurement) mitigates but does not eliminate this.
Breakout potential
Good if the wedge lands: 1071 extraction naturally expands to HMDA, CRA modernization, and general regulatory-report-from-loan-file tooling — a document-to-regulator pipeline for small financial institutions, all sold to the same compliance officer. Per the applied lesson set, this is a category (gov-mandate filing tools) the founder can serially exploit.
Final recommendation
CONDITIONAL GO — B-grade. The demand signal is the strongest structural kind (final federal rule, forced filers, cited), the build is squarely within solo capability, and the shape matches the founder's single proven win. It survives the kill attempt on buyer existence and willingness-to-pay grounds, but two cheap-to-resolve unknowns gate it: (1) does the final rule's coverage still include reachable small lenders, and (2) is the compliance-date calendar near enough to create 2026-27 urgency. Spend 2-3 days and $0 resolving both from the final rule text plus 10 compliance-officer conversations before writing code. If both check out, proceed to the paid Readiness Scan MVP; if coverage got gutted, kill without regret.
Next action
Read the final rule (federalregister.gov 2026-08494) today: extract the revised origination-threshold/coverage table, the compliance-date tiers, and the final data-point list; simultaneously start 10 outreach conversations with CDFI/community-bank compliance officers asking their 1071 plan and budget.