ACA Compliance & IRS 226-J Penalty Defense Engine

01 Thesis

Every Applicable Large Employer (ALE) — roughly, an employer that averaged 50+ full-time and full-time-equivalent employees in the prior year (Verified) — must file Forms 1094-C and 1095-C with the IRS every year reporting, employee-by-employee and month-by-month, whether it offered affordable minimum-essential coverage. When the IRS's systems decide an employer failed, it mails Letter 226-J proposing an Employer Shared Responsibility Payment (ESRP). For 2025 those penalties run $2,970 per full-time employee under §4980H(a) and $4,460 under §4980H(b); for 2026 the (a) penalty rises to $3,340 (Verified). A mid-size employer's proposed penalty is routinely six or seven figures, and a December-2024 law created the first-ever six-year statute of limitations on these assessments while extending the response window to at least 90 days (Verified).

We sell the outcome, not a tracking dashboard. For the recurring job, an internal AI engine ingests payroll and benefits data, builds the canonical full-time/affordability record, generates the Line 14/15/16 codes, validates against the IRS AIR schema, and we transmit and stand behind an Accepted filing. For the high-stakes job, when a 226-J (or a Letter 5699 / 5005-A / 972CG) lands, the engine reconstructs the year, reconciles the IRS's Form 14765 Employee PTC Listing line by line, and assembles a Form 14764 ESRP response that an enrolled agent or tax attorney certifies and files. The customer experiences a managed result — "the proposed penalty is wrong, here is the substantiated correct number, often $0" — not a tool they operate.

This is not a co-pilot. HR and finance teams under a 90-day IRS deadline do not want software to learn; they want the penalty handled by a professional who can practice before the IRS. The work is document-and-data synthesis against a fixed federal rulebook, so it decomposes into automatable steps with judgment concentrated at a few licensed chokepoints — the exact shape that lets revenue scale faster than headcount toward software-like margins.

02 Discovery rationale

This hour's broad scan swept HR/benefits compliance, federal tax penalty defense, logistics invoice recovery, pharmacy reimbursement, and pay-equity reporting for the recurring pattern this thesis rewards: a mandatory, deadline-bound, document-heavy filing against a fixed federal rulebook, where errors trigger large dollar penalties and the buyer already outsources adjacent work.

ACA reporting + 226-J defense won on four decisive, verified facts. First, the dollars are severe and rising — 2026 §4980H(a) is $3,340/FTE and the new six-year statute of limitations lengthens exposure. Second, a December-2024 statute gave employers a 90-day response window — enough operating runway for an outside vendor to assemble a defense (a 30-day window would have made the business impossible). Third, the Paperwork Burden Reduction Act removed the duty to mail 1095-Cs to employees but left the IRS-filing obligation fully intact — shifting the entire compliance burden onto exactly the data-accuracy step that triggers penalties. Fourth, a TIGTA review and the IRS's own figures (59,807 potential-ESRP cases identified for a single year) confirm the IRS systematically machine-matches ALE filings against marketplace subsidy data and issues penalty letters in waves — a recurring, predictable demand engine, not a one-off.

Freight/parcel audit was rejected: a mature, software-commoditizing market with a weak regulatory moat (it fails the Sam Altman test — better models help the incumbents' self-serve tools more than a service). Pharmacy DIR recovery scored well on "why now" but the highest-value piece (clawback appeals) verges on legal practice with a small-business buyer and thinner ACVs. Pay-transparency reporting is real but earlier and lower-stakes per filing. ACA penalty defense pairs the largest per-event dollars with a clean Circular 230 licensing chokepoint and a recurring filing annuity underneath it.

03 Candidate comparison

Five candidates generated this run. Scores are the author's 1–5 rubric averages (see §5 method), not external ratings.

04 Hard disqualifier check

No disqualifier fails. The watch-items are #11 (payroll incumbents own the data and could bundle defense) and pricing legality under Circular 230 §10.27 — both addressed in §17, §20, and §21.

05 Rubric scorecard

Six gates, scored 1–5.

06 Opportunity

The opportunity is two stacked revenue layers: a low-churn annual filing annuity priced per employee, and an episodic, high-ACV penalty-defense event triggered by IRS notices. The penalty layer's ROI is self-evident — a $20k defense fee against a $400k proposed assessment closes itself.

07 Evidence quality & source-claim matrix

08 Why now

Verified regulatory changes

• Bigger dollars. §4980H(a) rises to $3,340/FTE for 2026; (b) is $4,460 for 2025. Inflation indexing keeps penalties climbing, raising the value of getting filings right and defenses won.
• Six-year statute of limitations (new). Before ERIA the IRS arguably had unlimited time; now a defined six-year clock both bounds exposure and signals durable, ongoing enforcement employers must plan around.
• 90-day response window. The prior 30-day window left no room for an outside vendor to mobilize. Ninety days is exactly enough to reconstruct a year and assemble a substantiated response — the change that makes this a business.
• Paperwork Burden Reduction Act. Removing the mail-to-employee duty while keeping the IRS-filing duty pushes the whole burden onto data accuracy — the precise failure mode that produces 226-J letters.
• 10-form e-file threshold. Small ALEs that used to paper-file are now forced into AIR e-filing, where schema/business-rule rejections create new error surfaces.

Inferred capability changes

• Frontier LLMs now reliably normalize heterogeneous payroll exports (ADP, Paychex, Workday, UKG, gusto, spreadsheets) into a canonical monthly hours/enrollment record — historically the most expensive manual step.
• Code-generation and 14765 line-by-line reconciliation can be expressed as deterministic rules with LLM-assisted exception handling, compressing analyst time per return.

Unverified hypotheses

• That a renewed enforcement wave is imminent in 2026–2027 (timing of IRS letter batches is not publicly pre-announced) — treat as a tailwind, not a foundation.
• That mid-market ALEs are materially under-served on defense (vs filing) — to be confirmed via channel interviews in the pilot.

09 Customer & PMF

10 The outcome we sell

Measurable success metric: (A) AIR first-pass acceptance rate ≥ 99%; (B) average proposed-ESRP reduction (dollars abated ÷ dollars proposed) tracked per cohort, with a target of substantial reduction on the majority of substantiable letters. We never guarantee a specific dollar outcome with the IRS — that would be improper.

11 Internal AI engine architecture

12 AI-vs-human operations pipeline

AI and deterministic rules own the volume; trained operators handle data-completeness exceptions; the licensed professional owns every position taken with the IRS and is the only signer.

13 Operations as product

The operation is the product. Variance — not creativity — is the enemy, because variance is what produces a bounced filing or a missed deadline.

• SOPs & structured intake checklists per payroll system, with required-field manifests so a return cannot enter production with missing hours/enrollment data.
• Automated completeness checks that quantify data coverage before any analyst time is spent.
• Exception queues by type (missing months, ambiguous FT status, controlled-group questions) routed to the right skill level.
• Confidence scoring on every generated code; low-confidence codes auto-escalate to human review.
• Audit trails & version control on every record, code, and document — the same evidence file that defends the client also defends the firm.
• Gold-standard examples & red-team checks: adversarial test cases (rehires, COBRA, mid-year plan changes, leap-month edge cases) run against every model/logic change.
• Root-cause analysis & postmortems for every bounced filing or adverse 227 — the correction becomes a new rule or template, not a one-off fix.

Experts improve the machine; the machine, not heroics, carries the volume.

14 No-holes quality engine

• Two-key code generation: the LLM proposes codes; the deterministic rules engine independently derives them; mismatches block the return until reconciled.
• Schema gate: nothing transmits to AIR without passing the IRS XML schema and business-rule validators locally first — eliminating the most common "Accepted with Errors" outcomes.
• Penalty-exposure simulation: we re-run the IRS's own §4980H matching against the proposed filing to predict a 226-J before it can be issued, and fix the offer/affordability gap proactively.
• Citations-or-it-didn't-happen: every position in a 226-J response is tied to a record, a code definition, or a regulation; unsupported assertions are blocked from the packet.
• Licensed sign-off: no filing or IRS response leaves without a Circular 230 practitioner's review and signature.
• Deadline control tower: every 226-J's statutory response date is tracked with escalating alerts; the firm's worst failure mode (a blown deadline) is engineered against.

15 Pricing, pricing legality & unit economics

Pricing model

• Recurring filing (Outcome A): managed per-1095-C fee, tiered by volume (illustratively $5–9/form at scale; Inferred), with a modest annual base for eligibility tracking, sold with an acceptance guarantee — i.e., the outcome, not a software seat.
• Penalty defense (Outcome B): fixed engagement fee tiered by employee count and proposed-penalty magnitude (illustratively $7,500–$35,000; Inferred). This is the primary model and the Circular-230-safe default.

Why hourly is rejected

Benefits attorneys and consultants defend 226-J letters hourly ($300–600/hr), which caps scale, punishes efficiency gains, and misaligns incentives (the slower the work, the higher the bill). Fixed-fee outcome pricing lets our automation accrue to margin instead of to discounts.

Pricing-legality analysis (Circular 230 §10.27)

Circular 230 generally bars contingent fees for matters before the IRS, with exceptions including services connected to an examination of or challenge to an original return and a claim filed solely re: the determination of penalties assessed by the IRS (Verified). A 226-J response is a pre-assessment proposal, so a contingency on it is legally gray. We therefore price defense as a fixed fee by default. A success/results-based fee will be used only where outside counsel confirms a specific matter fits a §10.27 exception, and always disclosed in writing. Proposed 2025 amendments may narrow the contingent-fee ban further; we monitor and re-paper accordingly.

COGS per unit (illustrative, Unverified — validate in pilot)

Target gross margin: 60–70% at scale (Unverified). Revenue/FTE target: $400–700k (Unverified) — a blend of high-margin filing volume and high-ACV defense events. The decisive lever is the share of work the licensed professional touches; the engine must keep that to certification + true judgment only.

16 Nonlinear scaling plan

Revenue decouples from headcount because every resolved filing and letter feeds the learning loop: code logic, rebuttal templates, and the penalty simulator improve, shrinking both operator and licensed minutes per unit. Licensed professionals are the scarce resource, so the system is engineered to spend their time only on certification and genuine judgment. Figures are illustrative targets pending pilot instrumentation.

17 Moat & Sam Altman test

Why model improvement strengthens us: the costliest steps — normalizing messy multi-system payroll exports and drafting precise 14765 rebuttals — get cheaper and more accurate as models improve, expanding margin and throughput without adding licensed headcount. Our durable moat is not the model: it is (1) regulatory accountability — only Circular 230 practitioners can sign and represent, and clients buy that accountability; (2) the outcome guarantee backed by E&O, which a self-serve tool cannot offer; (3) a proprietary outcome dataset of AIR error codes and 227 resolutions that trains a better penalty simulator each season; and (4) channel lock-in with brokers and payroll bureaus.

Strongest commoditization threat: payroll/HRIS platforms (ADP, Paychex, Workday) already hold the source data and could bundle "filing + light defense." Mitigation: out-execute on mid-market defense depth (where platforms are thin and fear E&O), guarantee the outcome, and embed via the same brokers those platforms rely on — turning potential competitors into referral channels for the hard cases they don't want.

18 Buyer-specific go-to-market

The buyer requires trust and moves on triggers, so the motion is channel-led + founder outbound to trigger events, not waitlist/creator content.

19 Pilot design & early-demand-trap mitigation

• Pilot cap: ≤ 8 recurring-filing clients and ≤ 10 live 226-J defense engagements in the first season — deliberately constrained learning laboratories.
• Pilot profile: mid-market ALEs in 2–3 variable-hour verticals on a small set of payroll systems (e.g., ADP + Paychex + one HRIS) so the normalization engine hardens against real formats fast.
• Success criteria: ≥99% AIR first-pass acceptance; documented, substantiated reductions on the majority of defensible letters; cycle time and EA-minutes-per-unit trending down across the cohort.
• What we deliberately constrain: we refuse bespoke payroll-system one-offs and custom report formats that don't generalize; every manual workaround is logged.
• Convert fixes to product: recurring manual steps become connectors, rules, templates, or simulator improvements before we add the next vertical or payroll system.
• What would kill the idea: if EA review time refuses to fall with volume (judgment can't be concentrated), or if reasonable-cause/corrected-code responses rarely move proposed ESRPs, the margin and outcome theses fail — and we stop.

20 Competitive landscape

21 Regulation, compliance & licensing boundary

• Practice before the IRS: responding to and representing on an ESRP is restricted to Circular 230 practitioners; operators and AI may prepare but never sign or represent.
• Prohibited claims: no guaranteed dollar outcomes with the IRS; no "we eliminate all penalties"; no legal advice unless rendered by an attorney within scope.
• Pricing legality: fixed-fee default; success fees only where §10.27 clearly permits, vetted by counsel and disclosed (see §15).
• Privacy & security: employment, enrollment, and contribution data are sensitive PII; SOC 2 program, encryption in transit/at rest, least-privilege access, signed data-processing terms, and BAAs where any PHI is implicated.
• Audit logs: immutable record of every data input, generated code, human decision, and transmission — the same file defends the client and the firm.

22 Compact founding team & expert map

23 Exhaustive risk register

24 Tech stack & build plan

25 Metrics & KPIs

Also tracked: data-completeness score at intake, two-key mismatch rate, escalation-to-licensed rate, cycle time per deliverable, COGS per 1095-C and per engagement, evidence-completeness rate, defense-to-recurring conversion rate, pilot conversion rate.

26 What could kill this

• Judgment won't concentrate. If real returns force the EA/attorney into most of the work, margins never reach target and the business is just a staffing firm.
• Defenses don't move the number. If corrected-code and reasonable-cause responses rarely reduce proposed ESRPs, the headline outcome promise collapses.
• Enforcement goes quiet. A multi-year lull in 226-J issuance starves the high-ACV layer (recurring filing survives, but the growth story weakens).
• Incumbent bundling. A payroll giant ships credible bundled defense and undercuts on price using data it already holds.
• A single blown deadline or wrong filing that causes a real client penalty — reputational damage in a trust-driven, broker-referred market is severe.
• Pricing-legality stumble under Circular 230 that damages the licensed principal's standing.

27 90-day validation & launch plan

Evidence gaps to close in pilot: actual gross margin & revenue/FTE; real ESRP reduction rates; EA-minutes-per-unit trajectory; mid-market willingness to pay for both layers; channel referral economics.

28 Sources

• IRS — Information reporting by Applicable Large Employers (1094-C/1095-C)
• IRS — Employer Shared Responsibility Provisions (§4980H)
• IRS — IRM 8.7.21, Employer Shared Responsibility Payment (ESRP) under IRC 4980H
• IRS — Types of employer payments and how they're calculated
• BoomTax — What is IRS Letter 226-J? Complete guide (2026), with penalty amounts
• Trusaic — ACA compliance & IRS penalty defense, complete 2026 guide (June 2026 update)
• Bressler, Amery & Ross — IRS Letter 226-J: what it is and how to respond
• Frier Levitt — ACA reporting relief: 90-day window & six-year statute of limitations (ERIA)
• Pillsbury — IRS eases Form 1095-C furnishing burden for employers (PBRA)
• ADP SPARK — Expanded relief for furnishing Forms 1095-B/1095-C (IRS filing still required)
• Congress.gov — H.R.3797 Paperwork Burden Reduction Act
• Selerix — What triggers a Letter 226-J (e-file threshold & enforcement)
• Tax Notes — TIGTA: IRS lacking in review of ACA employer payments (59,807 potential-ESRP cases)
• eFileMyForms (Sovos) — ACA 1095 pricing ($1.49/form state filing; setup fee)
• 1095EZ — 1095-C filing pricing plans ($18–20 per employee)
• Trusaic — ACA Complete (track, prepare, furnish, file, defend)
• Points North — ACA Reporter (software + ADP Marketplace channel)
• ACA-Track — full-service ACA tracking & reporting
• Federal Register — Contingent fees under Circular 230 (§10.27)
• The CPA Journal — Treasury proposed regulations amending Circular 230 (2025)

ALE filing obligations, the 226-J / 14764 / 14765 procedure, 2025–26 penalty amounts, the ERIA 90-day window & six-year statute of limitations, the Paperwork Burden Reduction Act furnishing relief, the 10-form e-file threshold, the TIGTA case-volume figures, the competitor and pricing landscape, and the Circular 230 §10.27 contingent-fee rule are Verified via the sources above (retrieved 2026-06-30). Per-engagement fees, COGS, gross margin, revenue/FTE, automation rates, and ESRP reduction rates are author estimates labeled Inferred or Unverified in §7 and §15 and must be validated in pilots before decisive use. This is a hard-to-fool blueprint, not a guarantee.