What changed
FACT (per provided convergence input): proposed 45Z regulations tie per-gallon credit money to registration, emissions-rate certification, and recurring per-batch substantiation, and are currently at the proposed/hearing stage. FACT (per cited signals): headless agents can now manipulate Office documents server-side with a single binary (OfficeCLI), and agent platforms now complete multi-hour, multi-app deliverable work rather than chat (OpenAI). HYPOTHESIS: these two together make end-to-end automated assembly of credit-substantiation binders technically cheap for the first time.
Why now
The buying trigger is dated and imminent: producers must register and stand up documentation processes before final rules land, and every unsubstantiated batch is directly foregone credit revenue (FACT per input). HYPOTHESIS: no incumbent tax-software vendor targets small producers' batch-level workflow; incumbents serve large producers via consulting engagements. CAUTION: because rules are still proposed, exact substantiation formats may shift — the tool must be template-driven so format changes are config edits, not rebuilds.
Converging signals
(1) Government mandate: 45Z conditions real per-gallon credit dollars on registration + certification + recurring substantiation (input signals -1, -2). (2) Dev capability: OfficeCLI enables headless, server-side generation of the Excel/Word/PDF artifacts these packages consist of (github.com/iOfficeAI/OfficeCLI). (3) AI capability: long-horizon agents now produce finished multi-app deliverables, proving the 'document factory' pattern works for real knowledge work (openai.com/index/chatgpt-for-your-most-ambitious-work).
Customer pain
FACT (per input): small producers cannot staff recurring, deadline-driven document assembly, and unclaimed/unsubstantiated batches are direct lost revenue. HYPOTHESIS: the pain presents as either paying consultants thousands per engagement or leaving credits unclaimed; batch-level recordkeeping is the recurring grind that neither a CPA (annual cadence) nor an LCA consultant (one-time modeling) wants to own.
Who pays
Small and mid-size clean-fuel producers (biodiesel, renewable diesel, RNG, SAF blendstock) for whom each batch carries claimable credit value (FACT per input that they are the paying party; producer-type breakdown is HYPOTHESIS). Buyer is the owner/operator or plant controller — an SMB operational buyer, not enterprise procurement.
Solved today
HYPOTHESIS (consistent with how adjacent credits like RFS/LCFS are handled): spreadsheets plus a biofuels-savvy CPA for the tax filing, plus LCA/carbon-intensity consultants for the emissions-rate work; larger producers use consulting firms and carbon-accounting platforms. No source in the input documents current tooling, so treat as unverified.
Why current solutions are bad
Consultant engagements are priced for large producers and don't cover the recurring per-batch cadence; spreadsheets break under deadline-driven, repeated substantiation and produce audit risk. A missed or malformed package converts directly into disallowed credit dollars (severity is FACT-adjacent given credits are conditioned on substantiation; prevalence is HYPOTHESIS).
Proposed product
A hosted pipeline: producer connects or uploads batch production data + feedstock/emissions inputs once; every batch, the system generates a complete, versioned substantiation binder (Word/Excel/PDF via headless Office automation) — registration checklist, emissions-rate certification support docs, per-batch substantiation records, and a year-end package their CPA files from. Explicitly positioned as document assembly and recordkeeping automation, NOT tax advice — the CPA stays in the loop as the professional of record, which converts the trust problem into a channel.
MVP version
A template-driven generator: intake spreadsheet (batch volumes, feedstock, dates, CI inputs) → agent pipeline fills the substantiation binder templates → producer downloads a ZIP per batch. No integrations, no login-heavy SaaS; deliverable-quality output in week one using OfficeCLI-style tooling. Manually white-glove the first 3 customers ('we produce your binder') before automating self-serve.
30-day build
Interview 8-10 small producers via biodiesel/RNG trade associations and the public 45Z hearing-comment record to verify the bottleneck is document assembly (not CI modeling — this is the core kill-risk to test). Build the binder templates from the proposed-reg requirements. Land 2 pilot producers at a concierge price.
60-day build
Automate the pipeline end-to-end for pilot batches; recruit 2-3 biofuels CPAs/consultants as referral channel (they offload the grind, keep the filing). Publish a free '45Z substantiation checklist' as the lead magnet — mandate-driven search traffic converts well.
90-day revenue plan
5-10 producers on a monthly or per-batch plan. At $400-$1,200/mo per producer (HYPOTHESIS pricing anchored to a small fraction of credit value per batch), that is $2K-$10K MRR — real but modest; the wedge is the recurring cadence and year-end package lock-in into the first full tax cycle.
Distribution path
Direct, no-enterprise path: state/national biofuels and RNG trade association directories and newsletters, the NPRM public-comment list (commenters self-identify as affected producers — a public-records-sourced lead list, squarely in the founder's wheelhouse), biofuels CPA referral partnerships, and SEO on '45Z substantiation requirements'. Demonstrated-value sale: show a finished binder generated from their sample batch.
Pricing hypothesis
Concierge pilots at ~$1,500 setup + per-batch fee; steady state per-batch ($50-$150/batch) or flat monthly tier — mirrors his proven per-upload ELDT model. Anchor messaging to credit dollars recovered per batch, not software cost. All figures HYPOTHESIS pending pilot validation.
Technical difficulty
Low-to-moderate for the founder: data intake, deterministic template filling, headless Office generation, versioned storage — all proven patterns for him. The genuinely hard part (45ZCF-GREET emissions modeling / provisional emissions rates) should be explicitly OUT of scope v1: take CI values the producer's consultant already produced and assemble the substantiation around them.
Legal / regulatory risk
Material and must be managed: errors in substantiation packages could contribute to credit disallowance worth real money; unauthorized-practice concerns if positioning drifts toward tax advice. Mitigation: document-assembly positioning, CPA-in-the-loop, disclaimers, E&O insurance before scale. Also regulatory-change risk: rules are proposed, not final (FACT per input), so requirements may change — and credit-program political risk exists (HYPOTHESIS). This is the biggest structural weakness.
Platform dependency
Low. No app-store or API-platform gatekeeper; dependency is on the IRS rule text itself. OfficeCLI is open-source and replaceable with other document libs.
Founder fit
VERY HIGH pattern match with one honest caveat. This is exactly his proven ELDT shape: a federal mandate compels a party to register/certify/substantiate, and a solo operator builds the compliance layer and charges per transaction. His industrial-operations and recycling background gives real credibility with plant operators, and the lead-gen path runs through public records. Caveat: unlike ELDT there is no per-batch government portal to submit into — substantiation is largely audit-ready recordkeeping plus registration and the annual claim (HYPOTHESIS on exact mechanics pending final rules) — so the product is a document factory with a filing-support layer, not a pure portal-submission tool.
Breakout potential
Moderate-to-good: the same batch-data-to-binder engine extends to adjacent fuel-credit and registry regimes (LCFS reporting, RFS/RIN recordkeeping, 45V-style credits) and generally to any 'regulation compels recurring substantiation' niche — a repeatable playbook rather than a single product (HYPOTHESIS).
Final recommendation
PURSUE-WITH-GATE. The mandate-driven, per-transaction shape is the founder's proven playbook and the technical build is squarely solo-feasible, but two kill-risks are unresolved: whether document assembly (vs CI modeling) is the actual bottleneck, and whether buyers will pay before rules finalize. Spend 2 weeks on producer interviews and the NPRM comment record before writing product code; proceed only if 3+ producers confirm they would pay for a concierge binder now.
Next action
Pull the 45Z proposed-rule public comments, extract every small-producer commenter into a lead list, and book 8-10 interviews this week asking one question: 'who assembles your per-batch substantiation today, and what does it cost you per batch in fees or forgone credits?'