What changed
FACT (Federal Register 2026-10849, published 2026-06-01): DOL/OLMS published a combined final rule revising the LMRDA annual financial reporting forms — creating a new, longer Form LM-2 Long Form for the largest labor organizations, revising Form LM-2 for organizations at/above the $350,000 receipts threshold, and making a parallel revision to Form LM-3. FACT (Federal Register 2026-11799, published 2026-06-11): DOL issued a correction because a page was missing from the published Form LM-2 Long Form Instructions and the LM-2 Long Form and LM-2 forms contained formatting errors; the public is directed to www.dol.gov/olms (specifically dol.gov/agencies/olms/notice-lm-2-longform-rule) for correctly formatted forms and instructions.
Why now
The forms changed for the first time in years and the change is prospective, so every covered union's next annual filing must be prepared on a form its treasurer, bookkeeper, and accounting software have never produced before. INFERENCE: the correction notice itself — a federal agency shipping an instructions document with a page missing and formatting errors in the forms — is direct evidence that the new Long Form is complex enough to be error-prone at the source, which is exactly the condition under which filers pay for help. The window where confusion is highest and switching costs are lowest is the first filing cycle under the new form.
Converging signals
Three things meet at one point: (1) a final rule that redefines the form; (2) a defined, enumerable class of filers who cannot opt out (LMRDA-covered labor organizations above the $350,000 receipts threshold, with the largest on the Long Form); (3) an existing federal submission portal, the OLMS Electronic Forms System, that accepts the filing. HYPOTHESIS: a fourth signal is the correction notice, which implies a nontrivial rate of first-cycle filing errors and therefore rejected or amended filings.
Customer pain
HYPOTHESIS (not established by the provided sources — no complaint threads or job postings appear in the input): union treasurers and office managers must map a full year of receipts and disbursements into OLMS's specific schedules (disbursements to officers and employees, itemized receipts/disbursements above the reporting threshold, loans, investments, fixed assets), and a new, longer form invalidates whatever spreadsheet or prior-year template they used. The itemization schedules are the acute pain: they require transaction-level classification and per-payee aggregation that general accounting software does not produce in OLMS's shape. FACT: DOL itself could not publish the instructions completely and correctly on the first attempt.
Who pays
The labor organization itself, paid out of its administrative budget, or the CPA firm / labor-relations consultant that prepares LM filings on the union's behalf. The CPA/consultant is arguably the better buyer: they file for many locals, feel the pain N times per cycle, and buy tools without a board vote. INFERENCE, not established by the input: no evidence in the provided sources names any specific firm or price paid.
Solved today
OLMS provides free forms software and the Electronic Forms System for submission — FACT that a portal exists (named in the rule and the correction); INFERENCE that OLMS's own free tooling covers form entry and validation, as it historically has. Beyond that: an internal bookkeeper hand-keys the schedules, or an outside CPA prepares the return as a line item on the audit engagement.
Why current solutions are bad
HYPOTHESIS: the free OLMS tooling is a form-entry-and-validation surface, not a data-preparation surface. It will tell a filer that a total does not foot; it will not take a QuickBooks general ledger and produce the itemization schedules. The unsolved work is upstream of the portal — classifying and aggregating transactions into OLMS's schedule taxonomy. That is where a CPA's hours go and where a software wedge exists. This distinction is the entire thesis of the product and it is UNVERIFIED by the provided sources.
Proposed product
A ledger-to-LM-2 preparation tool. Ingest a general ledger export (QuickBooks, Xero, CSV) plus a payroll/officer-compensation export. Apply a classification engine — rules plus LLM-assisted categorization with human review — that maps every transaction to the correct LM-2 / LM-2 Long Form schedule and line, aggregates payees to test the itemization threshold, and flags the transactions a preparer must decide on. Output: a completed, internally consistent form package plus an exception report, ready to enter or import into OLMS EFS. Second product surface: a prior-year diff, showing what moved between the old LM-2 and the new Long Form so a repeat filer knows exactly which schedules are new.
MVP version
Narrow to the LM-2 Long Form (the largest filers, the newest form, the highest willingness to pay). Build: (1) a CSV/QBO general-ledger importer; (2) a schedule/line taxonomy transcribed from the corrected instructions on the OLMS site; (3) a rules-plus-LLM classifier with a review queue; (4) payee aggregation and itemization-threshold logic; (5) a foot-and-tie validator that reproduces the form's internal arithmetic checks; (6) PDF/print output plus a clean export. Do NOT attempt automated submission into EFS in the MVP — the founder's FMCSA TPR precedent shows submission automation is buildable, but it should be phase two after the preparation layer proves it sells.
30-day build
Read the corrected Form LM-2 Long Form and its instructions from dol.gov/agencies/olms/notice-lm-2-longform-rule and transcribe the complete schedule/line taxonomy into a machine-readable spec — this artifact is itself the moat and takes real work because the government's own version shipped with a missing page. In parallel, use the OLMS public disclosure database of prior-year LM filings to enumerate the actual filer universe by form type and receipts band, replacing the input's INFERRED '20,000–25,000 LM filers, a few thousand on LM-2' with a counted number. Identify the CPA firms that appear as preparers across many filings — those are the design partners.
60-day build
Build the importer, taxonomy, classifier, and validator against real prior-year filings pulled from the public disclosure database: for a union whose prior LM-2 is public, reconstruct it from a synthetic ledger and check that the tool reproduces the published numbers. Recruit three to five design partners from the enumerated CPA/consultant list, weighted toward firms that file for many locals. Price and sell on the first cycle before the product is finished.
90-day revenue plan
Charge per filing, invoiced at delivery, during the first filing cycle under the new form. Revenue is a function of the filing calendar, not of marketing: LMRDA annual reports are due ninety days after the end of the union's fiscal year, so a majority of filers cluster on a December fiscal year end and file in the first quarter. INFERENCE: that means the true revenue window for this product is the Q1 following the rule's prospective applicability, which may fall outside a 90-day horizon from today. Be honest that first revenue is calendar-gated, not effort-gated. Sell preparation engagements and multi-local CPA contracts now; recognize revenue when the cycle arrives.
Distribution path
Direct, cold, and evidence-led — which matches how the founder sells. The OLMS public disclosure database exposes, for every past filing, the union's name, address, receipts, and often the preparer. That is a complete, free, public list of every forced buyer and their size, and it lets the founder rank prospects by which schedules they will struggle with. Open with a free prior-year reconstruction: 'here is your last LM-2, remapped onto the new Long Form, showing the four schedules you have never filled out.' That is a demonstration of value, not a relationship sale. Secondary channel: the CPA firms that appear repeatedly as preparers.
Pricing hypothesis
Per filing, tiered by form: roughly $1,500–$3,000 for an LM-2 Long Form preparation, $500–$900 for a standard LM-2, with a per-seat annual license (roughly $4,000–$12,000) for CPA firms filing for many locals. HYPOTHESIS: this undercuts the incremental cost of CPA hours spent on the schedules. The input provides no evidence of any price actually paid by anyone for LM preparation — this pricing is constructed, not observed, and must be tested against a real design partner before it is believed.
Technical difficulty
Moderate and front-loaded. The hard part is not code, it is fidelity: transcribing the corrected form taxonomy exactly, and getting itemization-threshold and payee-aggregation logic right, because a filing that does not foot is worse than no product. There is a free, public regression corpus — every past LM filing in the OLMS disclosure database — against which the engine can be validated. That is an unusually good position for a solo builder. No submission-portal reverse engineering is required for the MVP.
Legal / regulatory risk
Low as to platform, real as to liability. Preparing a federal financial report that a union officer signs under penalty of perjury means an error in the tool becomes the customer's legal exposure. The product must be positioned as preparation software with a human preparer in the loop, not as an attestation. Errors-and-omissions coverage and clear terms are table stakes. The founder does not need a license to sell preparation software, so heavy_compliance is not flagged; this is ordinary product liability, and it is manageable.
Platform dependency
None meaningful. The counterparty is a federal agency with a statutory duty to accept the filing. There is no platform owner who can deplatform the product, no app-store review, and no API terms to violate. The only dependency is that OLMS keeps publishing the forms — which it is required to do.
Founder fit
Very high on shape, and the founder should be told exactly why. This is the FMCSA ELDT pattern: a federal mandate compels a defined class to submit a document to a government portal, and a solo operator monetizes per submission. He has shipped this before, against a harder portal. The public disclosure database gives him the same public-records advantage he already exploits. The gap between this and the ELDT product is that ELDT's filing is a certificate — a thin, schematized artifact — whereas an LM-2 is a full financial statement. The work here is accounting-domain classification, not portal plumbing, and that is a different and larger build than his prior win. He should not assume the ELDT playbook transfers wholesale.
Breakout potential
Contained. The filer universe is fixed and small (INFERENCE: a few thousand on LM-2, per the input's own estimate — unverified). This is a good business, not a large one: perhaps a few hundred thousand dollars of annual revenue at realistic penetration and pricing, recurring annually because the filing recurs annually. The genuine expansion is lateral, not vertical: the same ledger-to-federal-financial-form engine retargets to Form 5500, LM-3, LM-4, and state-level union and PAC financial reports. The taxonomy engine is the asset; LM-2 is its first market.
Final recommendation
CONDITIONAL — and the condition is a single afternoon's work, not a build. Do not start this as scoped. First, merge this item into 1535: it is the same opportunity, and the correction notice is not independent evidence of anything. Second, download the free OLMS forms software and attempt a filing with a sample ledger. If it imports a general ledger and generates the itemization schedules, kill the idea the same day. If it is a form-entry surface that leaves the classification work to the filer — which is what I expect but cannot show from these sources — then the wedge is real, and it is the upstream data-preparation layer, not the submission. Third, before writing code, pull the OLMS public disclosure database and count the actual LM-2 and Long Form filers. The input's '20,000–25,000 filers, a few thousand on LM-2' is explicitly labeled inference; the real number is free, public, and determines whether the ceiling justifies the build. Only then decide. The founder-fit is genuinely high and the forced-buyer structure is genuinely sound — but this brief rests on one primary source and a chain of plausible guesses, and I would rather say that plainly than dress the guesses up as a thesis.
Next action
Go to dol.gov/agencies/olms, download the free OLMS forms software and the corrected Form LM-2 Long Form instructions, and attempt to prepare a filing from a synthetic general ledger. Write down exactly where the free tool stops helping. That boundary — if it exists — is the product. If it does not exist, close this file.