Arbitrage Rebate Compliance Engine
Done-for-you, analyst-signed arbitrage rebate & yield-restriction computations for tax-exempt bond issuers and 501(c)(3) conduit borrowers — priced per issue, per computation period, with a deadline-met and computation-accuracy guarantee

1.Title

Arbitrage Rebate Compliance Engine — an AI-native, human-supervised post-issuance tax-compliance service that delivers completed, signed, audit-defensible arbitrage rebate and yield-restriction computations (not software the issuer operates) for the tax-exempt bond issuers and conduit borrowers who are legally obligated to compute and rebate excess arbitrage earnings to the U.S. Treasury under IRC §148(f).

Slug: arbitrage-rebate-compliance-engine • This is a federal tax-exempt-bond post-issuance compliance business (arbitrage rebate & yield restriction under IRC §148 and Treas. Reg. §1.148-3), explicitly distinct from prior municipal/finance runs in the manifest — it is not property-tax appeal, not unclaimed property, not ERISA 5500, not single audit, not cost segregation, and not the ASC 842 lease-accounting engine.

2.Final decision: Blueprint

3.Executive summary

When a state or local government — or a 501(c)(3) borrowing through a conduit issuer — sells a tax-exempt bond, the interest is exempt from federal tax only if the issuer follows the arbitrage rules: it may not earn and keep a profit by investing the borrowed proceeds at a yield higher than the bond's yield. Under IRC §148(f), any such excess ("arbitrage") must be rebated to the U.S. Treasury, computed and paid at least every five years and at final maturity, reported on IRS Form 8038-T, with payment due within 60 days of each computation date Verified [1][3][4]. The computation is technical (bond yield, future value of investment receipts, yield-restriction, and a thicket of exceptions), unforgiving (a missed deadline or a botched exception is a federal tax problem), and recurring for the life of every issue.

The market underneath this obligation is enormous and permanent: the U.S. municipal market is roughly $4.1 trillion outstanding with ~$513.6 billion issued in 2024 (+33% YoY), spread across more than 50,000 issuers and roughly one million distinct bonds, about 85% of them tax-exempt Verified [5][7]. Every tax-exempt issue with unspent or invested proceeds is a potential rebate computation, every year, until the bonds are retired.

The work is already outsourced to a stable set of specialist firms — BLX Group (since 1989), PFM, Hilltop Securities, Arbitrage Compliance Specialists, AMTEC, and The Arbitrage Group — which is the clearest possible proof that the budget exists and the buyer prefers to hand the obligation to an expert rather than run it in-house Verified [8][9][16][17][18][20]. The downside of getting it wrong is severe: a failure to comply with the arbitrage rules can result in loss of the tax-exempt status of the bonds under examination, typically resolved only by a closing agreement and a payment to the Treasury that is more favorable under voluntary correction than after an exam Verified [11][12].

4.Thesis

Arbitrage rebate is an excellent AI-native service unit because the atomic object — one bond issue for one computation period — is a discrete, evidence-grounded determination: given this bond's yield, issue price, and the dated cash flows into and out of each fund (project fund, reserve fund, refunding escrow, debt-service fund), what is the future value of the investment earnings, does any spending or small-issuer exception apply, is there a yield-restriction violation, and what is the resulting rebate or yield-reduction liability on Form 8038-T? An estimated ~75–85% of that work is structured extraction, deterministic financial math, and rule-checking — precisely where frontier models plus a calculation engine excel Inferred [4][14]. The residual is genuine judgment — does the 18-month or 2-year construction spending exception truly apply, how are transferred proceeds and the universal cap handled on a refunding, is a commingled fund allocated correctly — concentrated at a chokepoint owned and signed by a senior rebate analyst / tax professional.

Because the work is already outsourced, because the issuer cannot legally avoid the obligation, and because the customer experiences the company as an expert service (they receive a finished, signed report and a ready-to-file form), trust burden is low and the budget already exists Verified [8]. Pricing is naturally per issue, per computation/installment period — never hourly. As models improve at long-context document understanding and numerical reasoning, the autonomous share rises and cost-per-issue falls while the human concentrates on a shrinking exception residual (Sam Altman test: pass). The durable asset is not a model — it is the arbitrage operating system: the bond-document extraction templates, the §1.148 rule and exception library kept current, the SLGS/yield data feeds, the prior-period continuity ledger per issue, the QA that drives computation error toward zero, and the analyst sign-off that makes each report defensible to the IRS.

5.Discovery rationale

This run independently scanned the AI-native services terrain — public finance, tax/audit, insurance ops, healthcare administration, and regulated back-office finance — screening against the six gates and the evidence threshold, and explicitly checking the 57 prior blueprints in the manifest for overlap. An initial strong candidate, Work Opportunity Tax Credit (WOTC) capture, was rejected mid-research: the federal credit lapsed on December 31, 2025 and authority to claim it on post-2025 wages had not been reauthorized as of this run — an active regulatory blocker that makes a same-day "foolproof business" blueprint poorly timed Verified [23]. Discovery then converged on arbitrage rebate compliance for three reasons. First, demand is mandatory, statutory, and recurring: every tax-exempt issue with invested proceeds must be tested every period, for the life of the bonds, by force of federal law [1][4]. Second, the downside is catastrophic and the deadline is hard — loss of tax exemption, 60-day filing windows, and IRS examination authority — which is exactly the kind of regulated, high-stakes, document-heavy workflow where a service that removes the obligation commands willingness to pay [11][12]. Third — decisively — the incumbents are legacy human-analyst firms with a per-issue spreadsheet cost structure, and the entire prospect universe is public on EMMA, giving an AI-native entrant both a cost advantage and a distribution advantage. The scoping is deliberate: this is the federal arbitrage rebate & yield-restriction business, distinct from prior tax runs (R&D credit, cost segregation, property tax, sales-tax nexus, transfer pricing, ASC 842) and from the prior single-audit and ERISA 5500 filing engines.

6.Candidate comparison

Six AI-native service candidates were generated and scored (1–5, higher better) across demand evidence, gross-margin potential, MVP narrowness, licensing/regulatory safety, and whitespace vs. existing tools/vendors and the 57 prior blueprints. None of the prior blueprints duplicated the winner.

The winner leads on demand, margin, and MVP clarity; its only soft spots — licensing nuance (municipal-advisor registration boundary) and whitespace (an established specialist category) — are precisely the seams this blueprint is designed around: a computation-only scope that avoids MA-registration triggers, and an AI cost structure plus EMMA-driven distribution that legacy firms cannot match. Continuing-disclosure filing scored well and is retained as the natural second module, not a competing run.

7.CODE validation

C — Consumer / buyer trend

Two trends collide. (1) Rates revived rebate exposure. After a decade of near-zero short-term rates when invested proceeds rarely out-earned bond yields and rebate was routinely $0, the 2022–2024 rate surge means proceeds parked in SLGS, money funds, and reserve investments now frequently earn more than the yield on older low-coupon bonds — re-creating real positive arbitrage and yield-restriction liability that many issuers stopped monitoring Inferred [4]. (2) Thin government finance staffs and a heavy 2024–2025 issuance calendar mean more issues to monitor with fewer people, pushing the obligation outward to specialists Verified [5].

O — Opportunity

The underserved problem: mid-size and conduit issuers with recent construction-fund or refunding issues who lack in-house arbitrage expertise, may have let monitoring lapse during the ZIRP years, and now face revived liability and hard 5-year deadlines. Their alternatives are a legacy specialist firm (capable but slow and spreadsheet-priced) or in-house guesswork (dangerous). A faster, cheaper, equally-defensible AI-native service wins the long tail the big firms under-serve.

D — Demand

Buyers are visibly spending and procuring: state and local governments routinely issue public RFPs for "arbitrage rebate services"; bond counsel and municipal advisors customarily refer issuers to a rebate specialist at closing; GFOA publishes post-issuance compliance guidance telling issuers to put arbitrage monitoring in place; and a stable roster of specialist firms has sustained this as a standalone business for 35+ years Verified [8][18][19].

E — Economic sizing

With ~1 million bonds outstanding across 50,000+ issuers and ~85% tax-exempt [7], the population of distinct issues that could require periodic rebate testing runs into the hundreds of thousands; many are exempt via small-issuer or spending exceptions, but a large minority — every construction-fund issue, every funded reserve, every refunding escrow — needs ongoing computation. A conservative serviceable wedge: ~15,000 active issues needing annual monitoring × ~$2,000 blended per issue ≈ $30M of addressable annual spend in a single niche (e.g., conduit 501(c)(3) borrowers), against a far larger national post-issuance compliance budget Inferred [5][8]. A small share of a market this large supports a meaningful, high-margin business. Per-issue prices and the share owing rebate vary widely; figures are illustrative, not a forecast.

8.Rubric scorecard (six gates)

Total: 29 / 30. Anti-commoditization check (Gate 6 corollary): even if general models make a one-off computation self-serve, the durable wins are the maintained §1.148 exception library, the per-issue continuity ledger (each period builds on the last), the SLGS/yield data infrastructure, the IRS-exam feedback loop, and the analyst accountability the buyer is paying to offload — none of which a raw model provides.

9.Target buyer

10.Jobs-to-be-Done

• Functional: "Tell me whether this bond issue owes a rebate this period, give me the signed computation, and hand me a Form 8038-T I can file before the 60-day deadline."
• Risk/compliance: "Keep my bonds' tax-exempt status safe so an IRS exam never finds an arbitrage problem I didn't catch."
• Financial: "Don't let me overpay the Treasury — capture every spending exception, yield-reduction payment, and recoverable overpayment I'm entitled to."
• Emotional: "Stop the quiet dread that one of our older issues has a rebate liability nobody has been tracking."

11.Painful problem

Tax-exempt issuers carry a perpetual, easy-to-forget federal obligation: for the life of every issue, they must verify whether invested proceeds earned more than the bond yield and, if so, rebate the excess to the Treasury — computed at least every five years and at maturity, reported on Form 8038-T, paid within 60 days of each computation date Verified [1][3]. The mechanics are genuinely hard: bond yield, issue price, future-value of every dated investment receipt, the small-issuer exception (generally ≤ $5M, up to $10M/$15M for certain school bonds), and the 6-month / 18-month / 2-year spending exceptions each have precise tests Verified [4][15]. Miss a deadline and a late Form 8038-T draws a penalty ($25/day up to $1,000 per form) Verified [22] — but the real exposure is far larger: a substantive arbitrage failure found on IRS examination can cost the bonds their tax-exempt status, salvageable only through a closing agreement and a payment to the Treasury Verified [11][12]. Thinly-staffed finance offices cannot reliably track this across a portfolio of issues with different computation dates — so the obligation gets handed to a specialist, or it gets missed.

12.The outcome we sell

We do not sell a rebate-calculator tool, a co-pilot, or a dashboard the issuer operates. The issuer experiences an expert compliance service; the AI is the internal production engine.

13.First one-feature MVP wedge

14.Evidence summary

15.Claim table (Verified / Inferred / Unverified)

16.Source-claim matrix

17.Market & demand evidence

The market is defined by a federal mandate sitting on top of a vast, permanent asset base. SIFMA puts the municipal market at roughly $4.1 trillion outstanding, with $513.6 billion issued in 2024, up 33.2% year over year — a heavy issuance vintage that becomes tomorrow's monitoring workload Verified [5][6]. Across 50,000+ issuers and ~1 million distinct bonds (~85% tax-exempt), every tax-exempt issue with invested proceeds is a recurring compliance unit until retirement Verified [7]. The obligation is non-discretionary: §148(f) conditions the tax exemption itself on rebate compliance [1][4]. Demand is structural rather than cyclical — it is generated by law, refreshed by every new closing, and re-energized by the post-2022 rate environment that revived real arbitrage liability after a decade in which most computations were $0 Inferred [4].

18.Active buyer conversations

• Public RFPs. Governments and authorities regularly post procurement RFPs for "arbitrage rebate services" / "post-issuance compliance" on state and local bid portals — direct, dated evidence of budgeted spend and active vendor selection Verified [19].
• Bond-counsel & municipal-advisor referrals. Standard practice at closing is for counsel/advisor to recommend the issuer engage a rebate specialist and adopt written post-issuance procedures; the IRS information returns even ask whether such procedures exist Verified [10][19].
• GFOA guidance. The Government Finance Officers Association publishes best-practice materials urging issuers to implement arbitrage-rebate monitoring procedures — i.e., the buyer community is being told, repeatedly, to put this in place Verified [19].
• Trustee/corporate-trust touchpoints. The trustee bank that holds and invests the funds is a recurring conversation point and a natural referral and data partner.

19.Competitive landscape

No "no-competitors" delusion here: a mature roster of capable firms proves the budget. The opening is structural — every incumbent prices a senior human's spreadsheet time per issue, and none has an AI cost base or an EMMA-driven distribution machine to profitably reach the smaller/conduit long tail with a faster, equally-defensible product.

20.Competitor & budget validation

• Existing budget source: issuers already pay specialist firms (and, before that, bond counsel/advisors) for arbitrage rebate computations as a standalone line item; many sign multi-year monitoring engagements Verified [8][16].
• Incumbent alternatives: BLX, PFM, Hilltop, ACS, AMTEC, The Arbitrage Group — plus in-house finance staff and bond counsel for smaller issuers [8][18].
• Why current alternatives are insufficient: legacy firms are premium-priced and capacity-constrained for the small/conduit long tail; in-house staff lack expertise and deadline-tracking discipline; counsel does not want to own quantitative computation.
• Why we win: AI-native production drops marginal cost, enabling competitive per-issue pricing with faster turnaround and the same analyst-signed defensibility; public EMMA data gives precision targeting incumbents don't operate.
• Why not a clone: we are not another spreadsheet shop — the moat is the maintained rule/exception library, the per-issue continuity ledger, the IRS-exam learning loop, and a distribution engine built on public new-issue data.

21.Pricing evidence & proposed pricing

Incumbents do not post fees publicly; the established norm is a flat fee per bond issue per computation/installment period, tiered by complexity (number of funds, refunding vs. new money, draw-schedule length), with separate fees for arbitrage rebate vs. yield-restriction analysis vs. refunding verification Unverified (pricing level) [8]. Proposed pricing — never hourly:

Outcome/per-unit by construction. Any success-based component (e.g., a share of documented overpayment recoveries) is scoped conservatively and checked against the licensing boundary in §22 before use.

22.Regulatory & compliance considerations

The service operates inside a federal tax framework (IRC §148; Treas. Reg. §1.148-1 through -11) and an IRS enforcement regime (TEB examinations; VCAP) [11][12][14]. Three boundaries govern the business:

• Not legal advice / not a tax opinion. Arbitrage rebate computation is a quantitative compliance service; the firm computes and reports the number and prepares the form, but does not render the legal tax opinion on the bonds' tax-exempt status — that remains bond counsel's role. Engagement letters state this explicitly to avoid unauthorized practice of law.
• Municipal-advisor (MSRB) registration line. Providing advice to a municipal entity on the investment of bond proceeds can trigger SEC/MSRB municipal-advisor registration. The MVP scope is deliberately computation-only (we measure what was earned and what is owed); any forward-looking advice on how to invest proceeds is either excluded or delivered only after MA registration. This is the single most important compliance design choice and is reviewed with counsel before launch.
• Reliance & accuracy. Computations rely on issuer/trustee-provided data; engagement terms define data responsibilities, the computation-accuracy guarantee, professional-liability (E&O) coverage, and the rework/indemnity scope.

23.Licensing boundary

24.AI-native advantage

AI changes the unit economics, speed, and scalable reach of the service — not merely "use ChatGPT."

• AI tasks: parse the bond transcript and tax certificate to extract yield, issue price, dated maturities, fund structure, and computation dates; normalize trustee investment statements into dated cash flows; draft the computation narrative and the 8038-T; surface anomalies vs. prior period.
• Deterministic rule tasks: bond-yield and future-value math; small-issuer and 6-/18-/24-month spending-exception tests; universal cap and transferred-proceeds handling on refundings; yield-restriction computation.
• Human tasks: exception-qualification judgment, commingled-fund allocation calls, methodology sign-off, client-facing assurance.
• QC steps: independent re-performance/tie-out; prior-period continuity check; deadline gate; principal review above a liability threshold.
• Data inputs / output artifacts: bond docs + investment records in; signed report + 8038-T (or $0 report) + audit-defense file out.
• Never fully automated: the analyst sign-off and exception judgment — the defensibility the buyer is paying for.

25.Internal AI engine architecture

26.AI-vs-human operations pipeline

The default path is AI + rules; humans touch only the exception residual and the sign-off. Target automation share rises from ~70% at launch to ~88% by day 365 as the rule library and extraction templates harden.

27.Dynasty translation layer

• Buyer translation: the issuer's CFO/Finance Director pays to make a hard federal-tax deadline and a tax-exemption risk simply go away — outcome wanted: "my bonds stay tax-exempt and my 8038-T is filed on time, signed by an expert."
• Service translation: done-for-you computation. Customer receives a signed report + filed-ready form; AI handles extraction/math/drafting; humans qualify exceptions and sign.
• Workflow translation: intake → extraction → computation → exception review → QA → signed delivery → deadline-tracked renewal each period.
• Tooling translation: secure client portal/SFTP, document-AI + a deterministic calc engine, a rules/exception library, a per-issue continuity ledger, a deadline calendar, an audit-file vault — favoring off-the-shelf components before custom software.
• Sales translation: "We compute and file your arbitrage rebate, signed by a senior analyst, before your deadline — for a flat per-issue fee, faster than your current firm." Lead with a free portfolio diagnostic.
• Delivery translation: first 3 issues delivered semi-manually with heavy analyst oversight; automate extraction and computation as patterns repeat.
• Expansion translation: evolve into annual portfolio monitoring, continuing-disclosure (15c2-12) filing, refunding verification, and private-use/§141 monitoring — a full post-issuance compliance suite per issuer.

28.Anti-duplication analysis

• What exists: specialist rebate firms (BLX, ACS, AMTEC, etc.) and internal spreadsheet tools.
• Why not a copy: incumbents are human-analyst spreadsheet shops priced per senior-hour-of-effort; this is an AI-native production engine with a credentialed sign-off chokepoint and a public-data distribution machine.
• Narrow wedge: recent construction-fund / conduit 501(c)(3) issues — the most rebate-prone, least-served-by-premium-firms segment.
• Under-served segment: the small/mid-issue and conduit-borrower long tail that legacy firms find uneconomic to chase.
• Manual pain unsolved by tools: deadline tracking across a heterogeneous portfolio, exception qualification, audit-defense documentation, and continuity across periods.
• Differentiation: maintained §1.148 rule library + per-issue continuity ledger + IRS-exam learning loop + EMMA-driven outbound — none available from a raw model or a generic calculator.
• vs. prior manifest runs: distinct from cost segregation, R&D credit, property tax, sales-tax nexus, transfer pricing, ASC 842, single audit, and ERISA 5500 — different statute, buyer, deadline, and artifact.

29.Anti-commoditization analysis

If a future general model can do a one-off rebate computation from raw documents, the service still wins because the buyer is not paying for a number — they are paying to offload a federal-tax obligation onto an accountable expert. The durable assets are: (1) the credentialed analyst sign-off and E&O-backed defensibility the IRS and the issuer's auditors will accept; (2) the maintained exception/rule library that tracks regulatory and SLGS-rate changes; (3) the per-issue continuity ledger where each period depends on the last (switching costs); (4) the audit-defense file built and retained for the life of the bonds; and (5) the EMMA-driven distribution engine. A raw model provides none of these. Better models simply lower our cost and widen our margin.

30.Service delivery workflow

1. Engagement letter (computation-only scope; data responsibilities; guarantee).
2. Secure intake of bond transcript + trustee statements + draw schedule.
3. AI extraction + cash-flow normalization; rules/data retrieval for the issue.
4. Compute bond/investment yields, future-value rebate, yield restriction; run exception tests.
5. Analyst reviews exceptions & methodology; principal reviews liabilities above threshold.
6. QA re-performance + prior-period continuity reconciliation.
7. Deliver signed report + ready-to-file 8038-T (or $0 report) + audit-defense file.
8. Deadline calendar updated; renewal/next-period monitoring scheduled.

31.Operations as product

The operation is the product: variance is the enemy of both margin and defensibility.

• SOPs per issue type (new money, refunding, reserve-funded, construction-fund).
• Structured intake checklists + required-document lists; automated completeness checks.
• Exception queue with reviewer-assignment logic and confidence scoring.
• Per-issue continuity ledger and audit trail; version control on every computation.
• Gold-standard worked examples per exception; red-team checks on high-liability issues.
• Client-ready report + 8038-T templates; root-cause analysis and postmortem on any reviewer correction, missed deadline, or exam finding.

32.No-holes quality engine

• Independent re-performance: every computation tied out by a second path within tolerance before sign-off.
• Continuity gate: each period reconciled to the prior-period ledger; unexplained deltas block delivery.
• Deadline gate: no issue can sit un-scheduled; calendar enforces ≥ 15-day pre-deadline delivery.
• Confidence scoring: low-confidence extractions and edge-case exceptions auto-route to the analyst.
• Exam-feedback loop: any IRS inquiry triggers a root-cause review feeding the rule library.

33.What the human expert actually does

Blended human minutes per issue fall from ~80 at launch toward ~35 by day 90 as extraction and reconciliation automate; the irreducible core is exception judgment + sign-off.

34.Minimum viable offer

35.Fulfillment process

First 3 customers fulfilled semi-manually: AI does extraction and computation, but a senior analyst hand-checks every step and the founder watches each issue end-to-end to capture SOP-worthy patterns. By customer 5, intake and extraction are templated; by customer 10, the exception queue and continuity ledger are automated; computation and report drafting are AI-first with analyst sign-off throughout.

36.Tools & systems

• Secure client portal + SFTP (HTTPS, encryption at rest); document-AI for transcript/statement extraction.
• Deterministic arbitrage calc engine (bond yield, future value, yield restriction, exception tests) — the audited core, independent of any single LLM.
• Rules/exception library (§1.148) + SLGS/yield data feeds; per-issue continuity ledger; deadline calendar.
• Audit-file vault with retention for the life of the bonds; e-signature for reports.
• CRM seeded from EMMA new-issue data; engagement-letter + 8038-T templates.

37.Human-in-the-loop quality control

Two mandatory human gates per issue — exception qualification and signed review — plus a threshold-triggered principal review for high-liability computations and an independent re-performance tie-out. Humans never key the raw math; they own judgment and accountability. Every correction is logged and fed back into prompts and the rule library.

38.Nonlinear scaling & unit economics

Revenue/FTE scales because one analyst's signing capacity rises from a handful to ~15–20 issues/day as AI absorbs extraction, computation, and drafting. COGS tracked per issue from day one: model inference, document processing, storage/hosting, third-party data (SLGS/yield), human review, QA, E&O allocation, and sales follow-up. CAC payback target < 6 months given multi-period/portfolio retention. Conversion assumptions: portfolio-diagnostic → first paid issue ~25–40%; first issue → annual monitoring ~50%+; annual retention ~90%+ (continuity ledger + deadline dependency are sticky).

39.Distribution proof table

40.Sales & outreach plan

Lead with a diagnosis, not a demo. Use public EMMA data to identify recent construction-fund and refunding issues, send a personalized "arbitrage exposure snapshot," and convert to a free portfolio diagnostic → a paid single-issue computation → annual monitoring. Build bond-counsel and trustee referral partnerships in parallel for warm, high-intent flow. Compete on RFPs where price and SLA are explicit.

41.Founder-led content plan

Publish expert education for finance officers and advisors: why arbitrage rebate exposure returned with higher rates, how spending exceptions actually work, the 8038-T deadline mechanics, what an IRS TEB exam looks for, and overpayment-recovery basics. Position the founder/analyst as the plain-English authority; recycle top organic pieces into paid-ad and AEO assets.

42.First 30 days of content

10 educational posts

• "Arbitrage rebate is back: why higher rates revived a liability you forgot about."
• "The 60-day Form 8038-T deadline that can cost your bonds their tax exemption."
• "Small-issuer exception: do you actually qualify?"
• "6-month vs 18-month vs 2-year spending exceptions, explained."
• "What an IRS Tax-Exempt Bonds examination actually checks."
• "VCAP: fixing an arbitrage problem before the IRS finds it."
• "Yield restriction vs. rebate: the two arbitrage rules, untangled."
• "Construction-fund draws and reserve funds: where rebate liability hides."
• "Refundings, transferred proceeds, and the universal cap in plain English."
• "Are you overpaying the Treasury? Recoverable rebate overpayments."

3 diagnostic teardown formats

• Anonymized "exposure snapshot" of a recent construction issue.
• "Did this issuer owe rebate?" worked example with the math shown.
• "Deadline audit" of a sample portfolio's upcoming computation dates.

2 lead-magnet angles

• Free Arbitrage Exposure Snapshot (per issue, from public + minimal inputs).
• Post-issuance compliance & 8038-T deadline checklist (co-brandable with counsel).

1 webinar / live review

• "Arbitrage rebate in a higher-rate world" — live Q&A with bond-counsel guest, for GFOA-chapter audiences.

1 outbound diagnosis template

• "You closed a $X tax-exempt issue on [date]. Based on the structure, here are the three places rebate liability typically hides — want a free exposure snapshot before your computation date?"

43.Lead magnet & waitlist plan

Lead magnet: a free, no-obligation Arbitrage Exposure Snapshot — using public EMMA data plus a few issuer inputs, we estimate whether an issue is likely to owe rebate and flag the next computation deadline. Waitlist/CTA: "Get your free exposure snapshot." The artifact builds trust (shows expertise without doing the full paid computation), captures a strong pain signal (issuers who engage are worried about real exposure), and routes to a consult. A lead is sales-ready when it has a recent invested-proceeds issue and an approaching computation date.

44.Warm GTM plan

Work existing relationships first: bond counsel, municipal advisors, trustee officers, and any finance-officer network. Offer free diagnostics and scoped single-issue pilots. Convert diagnostic users and webinar attendees with a consultative review of their specific upcoming deadlines.

45.Targeted outbound plan

Build a precision list from EMMA: recent construction-fund and refunding issues, sized above a threshold, by 501(c)(3)/conduit and mid-size governmental issuers. Personalize each touch around the actual issue (amount, date, use of proceeds) and lead with the exposure snapshot — never a generic demo ask. Sequence: snapshot → diagnostic → single-issue computation → annual monitoring.

46.Answer-engine / search visibility plan

Own the questions finance staff ask AI and search engines: "do I owe arbitrage rebate," "Form 8038-T deadline," "small issuer exception," "spending exception 18 month." Publish authoritative, structured explainers and a calculator-lite so the company surfaces in ChatGPT, Perplexity, and Google answers, each with a diagnostic CTA.

47.Pilot design & early-demand trap mitigation

First cohort: cap at 8 issuers (ideally conduit 501(c)(3) borrowers with recent construction issues). The trap to avoid is taking unlimited bespoke historical clean-up work; pilots are scoped to standard single-issue current-period computations, with any complex multi-year look-back priced and queued separately. Weekly feedback captures where extraction or exception logic needed human rescue; those become SOPs, rules, and templates — not permanent manual labor.

48.Early-access feedback flywheel

• First pilot cohort: ≤ 8 issuers; early-access pricing in exchange for structured feedback and reference rights.
• Cadence: weekly review of every issue where a human had to intervene.
• Product feedback vs custom work: recurring extraction/exception gaps = product (fix the engine); one-off bespoke historical reconstructions = custom (price separately).
• Corrections → system: each analyst correction becomes a rule, prompt update, template, QA check, or retrieval source.
• Fix-before-expand: extraction accuracy and exception coverage must hit targets before opening more pilots.

49.Build-before-scale checkpoints

• After 5 issuers: harden intake checklists, required-document lists, and completeness checks.
• After 10 issuers: harden SOPs per issue type, the exception queue, reviewer checklists, and report templates.
• After 20 issuers: pause new pilots until COGS/issue, rework rate, escalation rate, and cycle time are measured and on target.
• Acceptable temporary manual workarounds: hand-keying an unusual trustee statement format. Non-scalable signals: every issue needing bespoke exception research, or analysts redoing computations from scratch.

50.7-day launch plan

• Days 1–2: engagement letter + computation-only scope reviewed with counsel; E&O quote; portal/SFTP stood up.
• Days 3–4: build the EMMA-sourced target list; draft the exposure-snapshot lead magnet.
• Days 5–6: publish 3 cornerstone explainers + the snapshot landing page; line up 2 bond-counsel referral conversations.
• Day 7: send first 25 personalized outbound snapshots; book first diagnostics.

51.30-day launch plan

• Run 20–40 diagnostics; convert 2–4 paid single-issue computations.
• Deliver first issues semi-manually; codify SOPs from each.
• Ship the deterministic calc engine v1 + extraction templates for the top trustee formats.
• Sign 1–2 referral partners (counsel/advisor/trustee).

52.90-day launch plan

• 8 pilot issuers live; automation share to ~80%; rework < 4%.
• Convert ≥ 50% of single-issue clients to annual monitoring.
• Launch the continuing-disclosure (15c2-12) add-on as the first expansion module.
• Measure COGS/issue, cycle time, escalation rate; decide scale-up.

53.Metrics & KPIs

54.Risks & mitigations

Top-level: regulatory mis-scope (MA registration), computation error/E&O exposure, data-dependency, and incumbent response. Each is addressed below and in the full register.

55.Exhaustive risk register

56.What could kill this

The two existential threats: (1) a regulatory mis-step — crossing the MA-registration line or producing a materially wrong computation that triggers an E&O/tax-exemption disaster and destroys trust in a referral-driven market; and (2) failure to build the moat fast enough — staying a thin AI-wrapper that incumbents copy before the data/continuity/distribution advantages compound. Both are mitigated by disciplined scope, mandatory human sign-off + tie-out, and relentless conversion of every issue into reusable system improvements.

57.Go/no-go reasoning

Every evidence-threshold element is met: a clear buyer, a painful mandatory problem, proof the problem exists (federal statute + IRS enforcement), proof buyers already spend (35+ years of specialist firms + public RFPs), active demand evidence, competitor/budget validation, a credible reason to win (AI cost base + EMMA distribution), a narrow MVP wedge, a service-first delivery path needing no large platform pre-revenue, a 60%+ gross-margin path, no unresolved fatal regulatory blocker (computation-only scope + analyst sign-off), and a believable distribution path. GO.

58.Final recommendation

59.Sources

1. IRS — About Form 8038-T (Arbitrage Rebate, Yield Reduction and Penalty) — link
2. IRS — Instructions for Form 8038-T (10/2021) — link
3. IRS — Instructions for Form 8038-T (computation dates & 60-day rule) — link
4. IRS — Phase I Lesson 5: Arbitrage and Rebate Overview — link
5. SIFMA — US Municipal Bonds Statistics — link
6. SIFMA — 2025 Capital Markets Fact Book — link
7. Investor.gov — The Municipal Securities Market (issuer & bond counts) — link
8. BLX Group — Arbitrage Rebate (specialist firm since 1989) — link
9. PFM Asset Management — Arbitrage Rebate Consulting — link
10. IRS — TEB Post-Issuance Compliance: Some Basic Concepts — link
11. IRS — Understanding the Tax-Exempt Bonds Examination Process — link
12. IRS — IRM 7.2.3 Tax Exempt Bonds Voluntary Closing Agreement Program (VCAP) — link
13. IRS — Publication 5091, Voluntary Compliance for Tax-Exempt and Tax-Credit Bonds — link
14. Cornell LII — 26 CFR §1.148-3 General arbitrage rebate rules — link
15. Ahlers & Cooney — An Overview of Spend-Down Rules Surrounding Tax-Exempt Bonds — link
16. Arbitrage Compliance Specialists (ACS) — Bond Compliance / Arbitrage Rebate Services — link
17. AMTEC — Arbitrage Rebate & Refunding Verification — link
18. HilltopSecurities — Post-Issuance Compliance — link
19. GFOA — Developing and Implementing Procedures for Post-Issuance Tax Compliance — link
20. The Arbitrage Group — Federal Arbitrage Rebate Requirements — link
21. IRS — Publication 5271, A Guide for Issuers of Tax-Exempt Bonds (arbitrage compliance) — link
22. Accountably — Form 8038-T Complete Filing Checklist (late-filing penalty) — link
23. Rockerbox — WOTC Outlook 2025 (credit status; basis for rejecting WOTC candidate) — link