AI-Native Service Business · Hard-to-Fool Blueprint

Business Personal Property Tax Compliance & Rendition Engine

A done-for-you service for multi-location operators that delivers one coupled outcome every year: every business-personal-property rendition/return filed accurately and on time in every county that taxes it, with the taxable value actively minimized (ghost assets scrubbed, assets correctly classified and depreciated) and every assessment notice reviewed and contested before it becomes a bill. AI is the internal production line; a registered property-tax consultant / CPA is the customer-facing professional of record.

Run: 2026-06-30 · Hour 05 · NN 01 Sector: State & local tax — business personal property (tangible personal property) compliance Buyer: Tax / fixed-asset / controller leaders at multi-county, asset-heavy operators Pricing: Per-return managed filing + value-savings success fee on minimization Outcome: On-time accepted renditions + minimized assessed value, audit-defensible

01 Thesis

In roughly 38 states, businesses owe an annual tangible/business personal property (BPP) tax on furniture, fixtures, machinery, equipment, computers, and — in 14 states — inventory; 36 states tax machinery & equipment specifically (Verified, Tax Foundation 2025). Unlike real-property tax, which the county assesses for you, BPP is self-reported: the owner must file a rendition/return in every taxing jurisdiction where it has assets, every year, by a hard deadline (April 15 in Texas, with a 30-day extension to May 15), listing assets at cost by acquisition year. Miss it and Texas imposes a 10% penalty on the tax (50% for fraud) and lets the appraiser assign an arbitrary, usually inflated, value — and flags the account for audit (Verified).

We sell the outcome, not a compliance dashboard. An internal AI engine ingests the customer's fixed-asset ledger and depreciation schedule, maps every asset to the correct jurisdiction, asset class, and cost-index/depreciation table, scrubs ghost assets (disposed equipment still on the books — averaging 15–30% of asset ledgers; Verified) that are silently inflating the tax, generates each county's exact form, and we file. When a notice of appraised value arrives that overstates the assets, the engine reconciles it and a registered consultant files the protest. The customer experiences a managed result — "every return is filed, your assessed value is as low as the law allows, and nothing slips a deadline" — not a tool they operate.

This is not a co-pilot. A controller running 40 restaurant locations across 6 states does not want software that surfaces 400 county deadlines for her staff to chase; she wants the filings done and the over-assessments fought, by someone who can be the agent of record before the appraisal district. The work is document-and-data synthesis against fixed statutory rules and county forms, so it decomposes into automatable steps with judgment concentrated at a few licensed valuation chokepoints — the shape that lets revenue scale faster than headcount toward software-like margins.

02 Discovery rationale

This run scanned several opportunity zones for regulated, document-heavy, deadline-driven administrative work that is commonly outsourced and decomposes cleanly: state & local tax compliance, logistics cost recovery, healthcare drug-pricing compliance, and manufacturing tax incentives. Five candidates were generated and three deep-validated (see §3). BPP compliance won on the combination of (a) verified, statutory, recurring demand — a mandatory annual self-report in ~38 states with explicit penalties for non-filing; (b) a clean recovery hook — the well-documented ghost-asset / over-reporting problem that lets us prove dollar savings on the very first cycle; (c) a real licensing chokepoint (Texas TDLR property-tax-consultant registration; CPA/attorney exemption) that both forms a moat and bounds the model legally; and (d) the fact that the largest incumbent just publicly validated the category by launching an AI-infused managed service in August 2025 — confirming demand and direction while leaving the under-served mid-market wedge open.

The decisive evidence was the asymmetry between how mechanical the work is (ledger → jurisdiction map → cost-index table → county form) and how painful it is at scale (hundreds of county deadlines, ghost assets quietly compounding the tax every year). That asymmetry is exactly where an internal AI engine plus a thin licensed-review layer can produce a done-for-you outcome at a margin a regional CPA firm doing it by hand cannot.

03 Candidate comparison

Five candidates generated this run; the top three deep-validated. Scores are the author's 1–5 rollup of the 15-factor screen (§5).

CandidateBuyerOutcome soldScoreEvidenceVerdict
BPP Tax Compliance & Rendition Engine Tax/fixed-asset/controller at multi-county asset-heavy operators Every BPP return filed on time in every county + minimized assessed value, audit-defensible 4.5 Strong SELECTED — statutory recurring demand, recovery hook, licensing moat, mid-market whitespace.
340B Discount-Capture & Compliance Engine Pharmacy / 340B program director at covered entities (DSH hospitals, FQHCs) Captured 340B savings + HRSA-audit-defensible program records 3.7 Strong demand Rejected — 48 identified TPA firms, 8 PE-owned, vertically integrated by PBMs/wholesalers; saturated, deep-pocketed incumbents; high commoditization & policy-volatility risk.
Parcel / Freight Invoice Audit & Recovery Logistics/ops leader at high-volume shippers Recovered carrier overcharges & late-delivery credits 3.0 Verified Rejected — commoditized (25–50% contingency standard), API platforms already automating it, no regulation moat (carrier contract terms, not law); fails Sam Altman / whitespace.
Manufacturing Utility Sales-Tax Exemption (Predominant-Use Study) Plant controller / tax manager at manufacturers Utility sales-tax exemption + refund of prior overpayments 3.2 Inferred Rejected — the defensible study typically needs an on-site engineering survey of equipment energy use; trips hard disqualifier #2 (physical/field labor).
Merchant Chargeback Representment Engine Risk/ops leader at e-commerce merchants Recovered revenue from won card disputes 2.9 Inferred Rejected — governed by card-network rules, not regulation; crowded, heavily automated, high commoditization; regulation-as-moat gate fails.

04 Hard disqualifier check

#DisqualifierStatusNote
1Customer-facing co-pilot / SaaS, not done-for-youPassWe file the returns and fight the assessments; customer hands over a ledger, receives an outcome.
2Requires physical labor / field crews / site visitsPassPure document/data work. (Note: a physical fixed-asset inventory is optional and is referred to a partner; our ghost-asset scrub is ledger-based.)
3Primary pricing is hourly / cost-plusPassPer-return managed fee + success fee on verified value reduction. No hourly billing.
4Cannot plausibly reach 50%+ gross marginUnclearPlausible as automation rises and review minutes fall; unproven — central pilot kill-metric (§15).
5Buyer cannot be identifiedPassTax manager / fixed-asset accountant / controller at multi-county asset-heavy operators.
6Workflow cannot be decomposedPassLedger ingest → jurisdiction map → classification → cost-index/depreciation → form-fill → file → notice reconciliation → protest.
7Fully automates regulated judgment without licensed reviewPassValuation positions & protests are signed by a registered property-tax consultant / CPA / attorney; AI never files a valuation opinion unattended.
8Substantially duplicative of a prior blueprintPassDistinct from the commercial-property-tax appeal engine (real property, contingency); see §20 & manifest similarity notes.
9Likely illegal / un-incorporable licensingPassLicensing (TX TDLR; analogous state rules) is explicitly designed into the chokepoint layer (§21).
10Core demand unverified / not inferablePassMandatory statutory annual filing + explicit penalties = Verified demand.
11Frontier models commoditize rather than strengthenUnclearAvalara/Ryan are productizing AI managed services now; moat must come from licensing + county-form coverage + outcome accountability, not model access (§17).
12Cannot be tested with a small bounded pilotPassOne multi-county customer, one filing season, measured savings — a clean bounded test.

No disqualifier fails outright. Two are Unclear (margin curve; commoditization by funded incumbents) and are carried into the rubric, risk register, and 90-day kill-criteria rather than hand-waved.

05 Rubric scorecard

Low trust burden
4.6
Low task-level judgment
4.2
High intelligence threshold
4.0
Regulation as moat
3.9
No physical labor
4.6
Sam Altman test
3.8

Aggregate ≈ 4.2 / 5. Strongest on low trust burden (BPP compliance is already routinely outsourced to CPA firms and SALT consultants) and no physical labor. Weakest on regulation-as-moat and Sam Altman — the rulebook is public and frontier models plus funded incumbents (Avalara, Ryan) can replicate the extraction layer; the durable edge is the licensed agent-of-record relationship, the breadth of encoded county forms/rules, and accountable outcome pricing.

06 Opportunity

~38
States that tax some business personal property V
36
States taxing machinery & equipment V
15–30%
Typical ghost assets in a fixed-asset ledger V
10%
TX penalty for failure to timely render (50% for fraud) V
Apr 15
TX rendition deadline (ext. to May 15) V
~9%
BPP share of total taxable market value in TX V
14
States that tax inventory (8 fully, 6 partial) V
Aug 2025
Avalara launches AI-infused property-tax managed services V

BPP compliance is large, fragmented, recurring, and quietly leaky. Asset-heavy multi-location operators file dozens to thousands of renditions across counties whose forms, depreciation schedules, deadlines, and exemptions all differ. Two dollars are on the table: the penalty/risk dollars from missed or late filings, and the over-assessment dollars from reporting ghost assets, mis-classed assets, and original cost long past its useful life. The second is a perpetual leak — once a ghost asset is on a rendition, it is taxed every year until someone removes it (Verified), which is precisely why a recurring engine that scrubs and defends value compounds savings.

07 Evidence quality & source-claim matrix

ClaimLabelSource / basisConf.Business impact
~38 states tax some BPP; 36 tax machinery/equipment; 14 tax inventoryVerifiedTax Foundation TPP research & data (2025)HighDefines serviceable footprint & jurisdiction complexity.
BPP is self-reported via annual rendition; TX deadline Apr 15, ext. to May 15VerifiedTX Comptroller Form 50-144; Travis/Tarrant CAD rendition pagesHighCreates the recurring, deadline-driven job we own.
10% penalty for failure to timely render; 50% for fraud; arbitrary value + audit for non-filersVerifiedTX Tax Code §22 / appraisal-district FAQs (Bexar, Gregg, etc.)HighQuantifies downside the customer pays us to prevent.
Ghost assets average 15–30% of fixed-asset ledgers; ~30% of orgs don't know what they own; reported & taxed; removal saves current + future yearsVerifiedRSM; BDO; Accounting Today; The Tax Adviser (Oct 2025)HighThe recovery hook that proves first-cycle ROI.
TX property-tax consultant: TDLR registration (40 hrs ed., sponsor, exam ≥70%, $50); CPA/attorney exemptVerifiedTexas TDLR Property Tax Consultants program pagesHighDefines the licensed chokepoint & partial moat.
Avalara launched AI-infused Property Tax Managed Services (AvaMPT), Aug 18 2025VerifiedAvalara newsroom; CPA Practice Advisor; Accounting TodayHighValidates category & direction; raises commoditization risk.
Incumbent landscape: Ryan, DMA, Weaver, Paradigm, Aprio, BDO/RSM, Avalara/CrowdReasonVerifiedFirm service pages; Avalara product pagesHighConfirms outsourcing precedent; shapes wedge (mid-market).
BPP ≈ 9% of TX total taxable market valueVerifiedTX-CCRI state budget & taxation noteMedIndicates material aggregate tax base.
Per-return managed fee + success fee on value reduction can reach 50%+ gross marginUnverifiedAuthor model; analogy to per-state return fees ($200–500+)LowCentral economics — pilot kill-metric.
Specialist review compresses to a few minutes per return at steady stateInferredInferred from rules-bound, template-driven nature of workLowMargin lever — must be measured in pilot.
Mid-market multi-location operators will switch from DIY/CPA to a done-for-you AI-native vendorUnverifiedNo direct buyer interviews this runLowDemand-quality assumption — validate before scaling.
Serviceable bottom-up TAM (firms × returns × fee)UnverifiedNot pinned to one audited figure this runLowSizing — 90-day research item.

Decisive selection rested only on Verified claims (statutory filing duty, penalties, ghost-asset prevalence, licensing, incumbent validation). Every economic assumption is Inferred or Unverified and routed to the risk register and 90-day plan.

08 Why now

Verified shifts

The largest compliance-automation vendor, Avalara, launched an AI-infused property-tax managed service in August 2025 — a market signal that done-for-you, AI-produced BPP compliance is now a fundable category, not a thesis. State legislatures continue to reform TPP taxes (raising de minimis exemptions, narrowing inventory tax), which churns the rulebook every year and rewards a vendor that keeps an always-current jurisdiction map (Verified: Tax Foundation reform tracking).

Inferred capability shifts

Frontier models now reliably parse messy fixed-asset registers (mixed CSV/Excel/PDF, inconsistent asset descriptions and acquisition dates), classify assets to jurisdiction-specific categories, and fill heterogeneous county forms — the exact extraction-and-mapping bottleneck that made BPP compliance labor-bound. (Inferred from current document-AI capability; to be proven on real ledgers.)

Unverified hypotheses

That mid-market operators are actively dissatisfied enough with DIY/CPA-firm handling to switch vendors, and that they will accept success-fee pricing on value reduction. (Unverified — buyer discovery is the first 90-day task.)

09 Customer & PMF

AttributeDetail
ICPAsset-heavy operators with locations in multiple BPP-taxing counties/states and 0–2 in-house property-tax specialists: multi-unit restaurant/retail/c-store chains, hospitality, healthcare & senior-care groups, manufacturers & distributors, equipment-rental/leasing, staffing/MSP with deployed equipment, telecom/data-center, logistics & cold-storage.
Economic buyerDirector of Tax / VP Tax / Controller / CFO (cost & risk owner).
Champion / userProperty-tax manager or fixed-asset accountant who currently chases county deadlines in spreadsheets.
Urgent triggerRendition season (Q1–Q2 deadlines); a missed-filing penalty or surprise inflated assessment; a new-location expansion; departure of the one person who "knew the counties"; an M&A asset-ledger integration.
Alternatives today(1) DIY in spreadsheets; (2) regional CPA firm by the hour; (3) self-serve software (Avalara/CrowdReason, TotalPropertyTax); (4) enterprise SALT firm (Ryan/DMA) — usually min. engagements that price out the mid-market; (5) do nothing & eat penalties + over-assessment.
Jobs-to-be-done"File every return on time in every county so we never eat a penalty or a default value"; "stop paying tax on equipment we scrapped years ago"; "make assessment notices someone else's problem"; "give me an audit-ready file if a county comes knocking."
Willingness to payInferred from existing per-return fees and contingency norms in property-tax consulting; the value-reduction success fee is self-funding when ghost-asset savings are real. Must be confirmed in pilot.

10 The outcome we sell

Deliverable

For each tax year: (1) every required BPP rendition/return prepared and filed (or filed by the agent of record) in every taxing jurisdiction by deadline, with proof of filing; (2) a documented value-minimization pass — ghost assets removed, assets re-classed, idle/obsolete assets flagged, applicable exemptions claimed (e.g., Texas Freeport/pollution-control where eligible); (3) review and, where warranted, protest of each notice of appraised value; (4) a maintained, audit-defensible workpaper file.

Acceptance criteria

All in-scope returns filed by deadline (zero late filings); assessed value at or below the prior-year baseline net of asset growth; every appraisal notice dispositioned (accept or protest) before the protest deadline; an audit packet producible within 24 hours per account.

Customer promise

"You will never miss a BPP deadline, you will not pay tax on assets you no longer own, and you will not face a county alone."

Exclusions / refund-rework

Excludes real-property tax, income/franchise tax, and litigation beyond administrative protest. If we miss a filing deadline we caused, we cover the resulting late penalty and re-file at no charge. Success fee applies only to verified, realized value reductions.

Measurable success metric: on-time filing rate (target 100%), realized assessed-value reduction vs. baseline, notices dispositioned before deadline, and audit-packet turnaround.

11 Internal AI engine architecture

1 · Intake

Fixed-asset register + depreciation schedule (ERP/GL export: NetSuite, SAP, Sage FAS, QuickBooks), prior-year renditions, location/asset roster, exemption certificates, and prior appraisal notices — via secure upload or read-only ERP connector.

2 · Normalization

Parse heterogeneous ledgers; standardize asset descriptions, acquisition dates, cost basis; dedupe; reconcile to GL totals; version every asset row with provenance.

3 · Retrieval / knowledge

Jurisdiction registry: per-county form, deadline, depreciation/cost-index table, asset-class definitions, de minimis & exemption rules, e-file portal specifics — kept current as a structured, versioned dataset.

4 · AI workbench

Map each asset → jurisdiction + asset class; apply the correct depreciation/index schedule; flag ghost/idle/obsolete assets and probable mis-classifications; draft each county form; draft notice-reconciliation memos.

5 · Deterministic rules

Deadlines, extension logic, penalty math, de minimis thresholds, exemption eligibility tests, and cost-index lookups run as code — never as model guesses.

6 · Human chokepoint

Registered property-tax consultant / CPA reviews valuation positions, signs renditions where an agent of record is required, and approves/files every protest. Operators clear exception queues.

7 · QA

Pre-file completeness & reconciliation checks; ghost-asset removal sign-off; cross-year variance review; second-set-of-eyes on any return above a value threshold.

8 · Delivery

File via county portal / mail / agent-of-record submission; deliver proof-of-filing, the value-reduction summary, and the audit packet to the customer.

9 · Learning loop

Assessor adjustments, protest outcomes, audit findings, and form rejections feed back to improve classification, depreciation, and exemption logic and the jurisdiction registry.

10 · Model portability

Model-agnostic extraction/classification interface; swap or ensemble frontier models per task; deterministic layers and the jurisdiction registry are model-independent assets.

12 AI-vs-human operations pipeline

AI

Ingest & normalize fixed-asset ledger; reconcile to GL.

AI

Map assets to jurisdictions & asset classes; flag ghost/idle/obsolete.

Rules

Apply depreciation/cost-index tables, deadlines, de minimis, exemption tests.

AI

Generate each county's rendition form & supporting schedule.

Operator

Clear exception queue: unmatched assets, low-confidence classes, missing data.

Licensed consultant

Approve valuation positions; sign as agent of record where required; authorize protests.

Rules

Pre-file completeness, reconciliation & penalty checks; lock the package.

AI

File via portal/mail; capture proof; reconcile incoming appraisal notices.

Customer

Receives proof-of-filing, value-reduction summary, audit packet.

Humans are concentrated at two chokepoints — operator exception-clearing and licensed valuation/protest sign-off. Everything upstream and the filing mechanics are AI- or rules-owned.

13 Operations as product

  • SOPs per workflow: ledger intake, jurisdiction mapping, ghost-asset scrub, form generation, notice reconciliation, protest filing — each a versioned runbook.
  • Structured intake checklist + automated completeness check before any account enters production.
  • Jurisdiction registry as the core asset: every form/deadline/table/exemption versioned, dated, and source-linked; changes reviewed before a cycle.
  • Exception queues with confidence scoring: assets the model can't confidently classify or value route to operators; only true valuation calls route to the consultant.
  • Reviewer assignment logic by jurisdiction, account value, and risk tier.
  • Audit trail & version control on every asset row, form, and filing — the audit packet is a byproduct, not extra work.
  • Gold-standard examples per county form; red-team checks for over-/under-reporting and missed deadlines.
  • Root-cause & postmortem on every late filing, rejected form, or lost protest, converted into a rule or registry update.

The product is the production system: experts improve the machine and handle the hardest valuation calls; the machine handles the volume and never forgets a county.

14 No-holes quality engine

Deadline integrity

Every in-scope jurisdiction has a tracked deadline with automatic extension logic and escalation; a daily reconciliation proves "every account, every county, accounted for." Missing-deadline = P0 alert.

No over-reporting

Ghost-asset and obsolescence checks run before filing; assets above useful life flagged; the goal is the lowest defensible value, never an aggressive or fabricated one.

No under-reporting / fraud risk

Reconciliation to the GL and prior-year baseline prevents omitting taxable assets; large downward swings require consultant sign-off and documented rationale (defends against the 50% fraud penalty).

Hallucination control

Depreciation, penalty, and exemption math are deterministic; model outputs are constrained to the jurisdiction registry; every value cites its source asset row and table.

Form correctness

Each county form validated against a stored schema/gold example; e-file portal rejections feed the learning loop.

Audit readiness

Every filing carries linked workpapers; a county audit is answered from the packet, not a scramble.

15 Pricing, pricing legality & unit economics

Model

Primary: per-return managed compliance fee (tiered by jurisdiction complexity and asset count), billed per filing cycle. Secondary: a success fee on verified, realized assessed-value reductions (e.g., a share of first-year tax savings from ghost-asset scrubs and successful protests). Why not hourly: hourly billing caps margin, punishes the customer for messy ledgers, and rewards slowness — the opposite of an engine whose whole thesis is that each return gets cheaper to produce.

Pricing legality

Per-return fixed fees are unrestricted. Contingency/success fees for property-tax representation are common industry practice but are state-regulated: in Texas, valuation representation must be performed by a TDLR-registered property-tax consultant (or exempt CPA/attorney), and consultant conduct is governed by TDLR rules. Where contingency representation is restricted or the work is pure return preparation, we default to fixed per-return + a fixed value-review fee. Engagement letters disclose fee basis, scope, and the licensed agent of record. (Verified licensing basis; specific contingency rules validated per state in §27.)

Illustrative unit economics — single mid-complexity return (Unverified, model)

COGS driverEst. per returnNote
Model inference (ingest, classify, draft)$0.30–1.50Falls as prompts/models optimize.
Document processing / storage / hosting$0.20–0.60Per-account amortized.
Operator exception-clearing$3–10Target 4–8 min at steady state; messy first-year ledgers higher.
Licensed consultant review/sign-off$2–8Only valuation positions & protests; most returns flow through.
e-file / portal / mail / filing$0.50–3Some counties paper-only.
QA + support + rework reserve$1–4Rework target <5%.
Total COGS / return~$7–27Against a target price well above this for 50%+ margin.

Targets: gross margin 55%+ at steady state; revenue/FTE $400k–700k as automation rises. All figures are author estimates (Unverified) and are the central pilot kill-metrics.

16 Nonlinear scaling plan

LeverLaunch+90 days+1 year
Automation rate (returns flowing without operator touch)30–45%55–65%75–85%
Operator minutes / return20–3510–184–8
Consultant touch (% of returns)~100% (trust-building)25–40%10–20% (valuation/protest only)
Returns / operator / day10–2030–6080–150
Gross margin0–30%40–55%55–65%

Revenue decouples from headcount because the jurisdiction registry and classification models are shared fixed assets: the 200th county encoded serves every customer with assets there, and the second year of any account is far cheaper than the first (the ledger is already normalized, ghost assets already scrubbed). Margin expands as automation rate climbs and consultant touch falls to valuation chokepoints only. (All Inferred/Unverified; the curve is the thesis under test.)

17 Moat & Sam Altman test

Does model improvement strengthen or commoditize us? Better models make ledger parsing, asset classification, and form-fill cheaper and more accurate — directly expanding margin and throughput. The engine is architected model-agnostic, so we ride the curve. Score: 3.8.

What actually defends the business (because the rulebook is public and incumbents are funded): (1) the licensed agent-of-record relationship and accountable outcome — a customer fires a vendor that misses a deadline, not one whose model is 2% better; (2) breadth and currency of the encoded jurisdiction registry (thousands of county forms, tables, deadlines, exemptions), which is expensive to build and maintain and improves with every cycle; (3) switching cost from a normalized, ghost-scrubbed multi-year asset history living in our system; (4) proof-of-savings track record per vertical.

Strongest commoditization threat: a funded incumbent (Avalara's AvaMPT, Ryan, DMA) productizes the same AI managed service for the mid-market and bundles it with an existing ERP/tax footprint, compressing price before we reach registry breadth. Mitigation: win on accountable outcome + verticalized savings proof + speed in an under-served segment the enterprise firms decline; treat registry coverage as the priority asset.

18 Go-to-market

Buyer behavior (finance/tax leaders, trust-sensitive, deadline-driven) points to specialist-led outbound plus channel, not waitlist/creator motions.

Why this GTM

Mid-market tax leaders buy from credible specialists on referral and at trigger moments; they will not self-serve a compliance obligation with penalty exposure.

First 50 prospects

Multi-state restaurant/retail/c-store groups, senior-care & healthcare networks, equipment-rental firms, and manufacturers/distributors with locations across TX/FL/VA/CA and other BPP states.

Trigger events

Rendition season; a missed-filing penalty or inflated default assessment; new-location/M&A expansion; departure of the in-house property-tax person; ERP migration.

Outreach wedge

Free "Ghost-Asset & Exposure Scan": send us last year's renditions + fixed-asset ledger; we quantify likely over-assessment (ghost/obsolete assets) and any missed-jurisdiction risk — a concrete dollar number that funds the engagement.

Channel partners

Regional CPA firms (white-label overflow), ERP/fixed-asset implementers (Sage FAS, NetSuite), franchise networks, equipment lessors, and SALT-light advisory firms.

Conversion path & metrics

Scan → single-state managed pilot (one cycle) → full multi-state program → ERP integration. Track scan-to-pilot rate, pilot-to-program rate, on-time filing rate, realized savings.

Credibility asset: a registered property-tax consultant on the team plus a published, verifiable savings case study per vertical. Expected sales cycle: weeks-to-a-quarter, accelerating near deadlines.

19 Pilot & early-demand-trap mitigation

Pilot cap

Max 5–8 accounts for the first filing season, deliberately spanning only 2–4 states so the jurisdiction registry hardens before breadth.

Pilot customer profile

One multi-state operator with messy ledgers + several single-state accounts to stress the engine across complexity.

Success criteria

100% on-time filing; documented ghost-asset savings ≥ engagement fee; operator minutes/return trending down cycle-over-cycle; consultant touch falling below 100%.

Manual workarounds — tracked

Log every hand-fix (a county form the engine can't yet generate, a one-off exemption). Each becomes a registry entry or rule before scaling, never a permanent human patch.

What we refuse

No bespoke real-property or income-tax work; no "just file it however" requests that don't improve the engine; no jurisdiction we can't yet encode defensibly.

What kills the idea

If operator minutes/return don't fall across cycles, or ghost-asset savings are too small/contested to fund fees, or consultant review can't drop below ~100% without quality loss — margins won't clear and we stop.

20 Competitive landscape

CategoryExamplesGap we exploit
Enterprise SALT / property-tax firmsRyan, DMA (DuCharme McMillen), Altus, Paradigm Tax GroupHigh minimums & bespoke service; mid-market is unserved or price-gouged.
Self-serve compliance softwareAvalara Property Tax / CrowdReason TotalPropertyTax, othersStill customer-operated; we sell the outcome, not a tool to run.
AI managed services (new)Avalara AvaMPT (Aug 2025)Direct threat & validation; we differentiate on mid-market focus, accountable savings, speed, vertical proof.
Regional CPA / advisory firmsWeaver, Aprio, BDO, RSM, local CPAsHourly, capacity-bound, often treat BPP as an afterthought; we are purpose-built & cheaper at volume.
In-house tax/fixed-asset teamsSpreadsheets + ERP fixed-asset moduleDeadline-chasing, ghost assets unscrubbed; we remove the burden & the leak.
Do nothingPenalties, default inflated assessments, perpetual over-payment on ghost assets.

21 Regulation, compliance & licensing boundary

What AI/operators may do

Ingest ledgers, classify assets, apply depreciation/cost-index tables, draft forms, reconcile notices, prepare workpapers, and file ministerial returns.

What licensed professionals must do

Render valuation opinions, act as agent of record before an appraisal district, and file protests — performed by a TDLR-registered property-tax consultant (TX) or analogous state-registered consultant, or an exempt CPA/attorney.

Licensing facts V

TX TDLR property-tax-consultant registration requires 40 hrs education, a Senior-PTC sponsor, exam ≥70%, and fees; active TX CPAs and attorneys are exempt. We staff/contract registered consultants per state.

Prohibited claims

No guarantee of a specific assessment outcome; no aggressive under-reporting; we never advise omitting taxable assets. Success fee only on verified, realized reductions.

Privacy & data

Read-only ERP access, encryption, access controls, audit logs; financial data handled under a DPA. Customer's authorized officer signs renditions where the form requires owner certification.

Per-state expansion gate

Before entering a state, confirm consultant-licensing & contingency-fee rules and encode them into the engagement template and registry.

22 Founding team & expert map

RoleWhy neededFT / fractionalFirst hire
Property-tax domain lead (registered consultant or SALT CPA)Agent of record, valuation sign-off, credibilityFull-timeFounder / first hire
Ops leadOwns SOPs, exception queues, filing calendarFull-timeMonth 1
AI/automation engineerIngest, classification, jurisdiction registry, form-fillFull-timeFounder / first hire
Additional registered consultants (per state)Multi-state agent-of-record & protest coverageFractional → FTAs states added
Sales / channel leadOutbound + CPA/ERP partnershipsFractional → FTPost-pilot
QA ownerPre-file checks, audit-packet integrityFractional (ops doubles early)As volume grows

23 Exhaustive risk register

1 · Margin curve never materializes (review minutes stay high)
Likelihood: Med-HighImpact: CriticalInferred

If messy real-world ledgers keep operator + consultant minutes high, gross margin never clears 50%. Mitigation: measure minutes/return obsessively in pilot; convert every recurring fix into a rule/registry entry; price first-year cleanup separately from steady-state filing. Owner: Ops + AI lead. Leading indicator: minutes/return trend cycle-over-cycle.

2 · Funded incumbent commoditizes the mid-market (Avalara AvaMPT, Ryan)
Likelihood: Med-HighImpact: HighVerified threat

Avalara already launched AI property-tax managed services (Aug 2025). Mitigation: compete on accountable savings, mid-market focus, vertical proof, and speed; prioritize registry breadth and switching-cost data; avoid head-to-head enterprise RFPs early. Owner: Founder. Leading indicator: incumbent mid-market pricing & win/loss.

3 · Buyers won't switch from DIY/CPA without a crisis
Likelihood: MedImpact: HighUnverified

Mitigation: lead with the free ghost-asset scan (a dollar number), target trigger events (penalty, expansion, staff departure), white-label through CPA channel. Owner: Sales. Leading indicator: scan-to-pilot conversion.

4 · Ghost-asset savings are smaller or more contested than assumed
Likelihood: MedImpact: HighInferred

If real ledgers are cleaner than the 15–30% literature, the recovery hook weakens. Mitigation: don't over-index on success fee; per-return compliance value (deadline integrity, audit defense) stands alone. Owner: Domain lead. Leading indicator: measured savings per pilot account.

5 · Missed deadline caused by us → penalty + reputational hit
Likelihood: Low-MedImpact: HighInferred

Mitigation: deadline integrity as a P0 system with daily reconciliation, redundant alerts, and extension automation; we cover penalties we cause. Owner: Ops. Leading indicator: deadline-reconciliation exceptions.

6 · Under-reporting / fraud-penalty exposure from over-aggressive scrubbing
Likelihood: LowImpact: HighVerified penalty (50%)

Mitigation: lowest-defensible-value policy; consultant sign-off + documented rationale on large downward swings; GL reconciliation prevents omissions. Owner: Domain lead. Leading indicator: audit findings / assessor pushback rate.

7 · Jurisdiction registry drift (forms/tables/rules change yearly)
Likelihood: HighImpact: MedVerified churn

States reform TPP taxes and counties revise forms annually. Mitigation: registry is versioned with a pre-season refresh sprint; change-detection on portal/form sources. Owner: AI lead. Leading indicator: form-rejection rate.

8 · State licensing / UPL / contingency-fee restrictions vary
Likelihood: MedImpact: Med-HighVerified (TX)

Mitigation: per-state licensing gate before entry; registered consultants / exempt CPAs as agent of record; fixed-fee fallback where contingency restricted. Owner: Domain lead + counsel. Leading indicator: legal review sign-off per new state.

9 · ERP/ledger data quality blocks automation
Likelihood: Med-HighImpact: MedInferred

Mitigation: robust normalization layer; structured intake checklist; price first-year cleanup; offer optional partner-led physical inventory for chronic ghost-asset cases. Owner: AI + Ops. Leading indicator: normalization exception rate.

10 · Concentration / seasonality (Q1–Q2 deadline crush)
Likelihood: HighImpact: MedInferred

Most renditions cluster around spring deadlines, straining capacity. Mitigation: automation removes the labor peak; stagger onboarding to off-season; add notice-protest work (mid-year) to smooth revenue. Owner: Ops. Leading indicator: peak-season utilization.

11 · Legislative erosion of the BPP base (states keep exempting TPP)
Likelihood: Med (slow)Impact: MedVerified trend

Reform raises de minimis exemptions and narrows inventory tax, shrinking the footprint over time. Mitigation: diversify across the ~38 states & verticals; expand adjacent (notice defense, fixed-asset hygiene); treat as a multi-year, not existential, risk. Owner: Founder. Leading indicator: count of taxing jurisdictions YoY.

12 · Liability for an over-assessment we failed to contest
Likelihood: Low-MedImpact: MedInferred

Mitigation: every notice dispositioned before deadline with documented rationale; E&O insurance; clear scope in the engagement letter. Owner: Domain lead. Leading indicator: notices undispositioned at deadline (target zero).

24 Tech stack & build plan

Stack

Python services; frontier LLM(s) behind a model-agnostic extraction/classification interface; document parsing (PDF/Excel/CSV); a structured jurisdiction registry (Postgres, versioned); deterministic rules engine for deadlines/penalties/depreciation/exemptions; secure object storage with audit logging; ERP connectors (NetSuite, Sage FAS, QuickBooks) + secure upload; e-file/portal automation per county; internal ops console with exception queues & reviewer assignment.

Build sequence

(1) Ledger ingest + normalization + GL reconciliation. (2) Jurisdiction registry for beachhead state(s) — forms, tables, deadlines, exemptions. (3) Classification + ghost-asset flagging + deterministic depreciation. (4) Form generation + pre-file QA. (5) Filing + proof capture. (6) Notice ingestion + reconciliation + protest workflow. (7) Learning loop + multi-state registry expansion.

No "use agents" hand-waving: discrete services with explicit inputs/outputs, deterministic math in code, model calls only for extraction/classification/drafting under human review.

25 Metrics & KPIs

Throughput

Returns/operator/day; accounts/cycle.

Cycle time

Ledger receipt → filed; notice → dispositioned.

On-time filing rate

Target 100%; zero self-caused late filings.

Realized value reduction

$ saved vs. baseline (ghost-asset + protest).

Automation rate

% returns with no operator touch.

Consultant touch rate

% returns needing valuation sign-off.

Rework / quality-failure rate

Target <5%; form rejections trending down.

COGS/return & gross margin

Margin target 55%+ at steady state.

Revenue/FTE

Target $400k–700k as automation rises.

Notices dispositioned before deadline

Target 100%.

Audit-packet turnaround

Target <24h per account.

Scan→pilot→program conversion

Funnel health.

26 What could kill this

  • Margins never clear 50% because review minutes stay high on messy ledgers — the fastest, most likely killer.
  • A funded incumbent (Avalara/Ryan) bundles an AI managed service into the mid-market at a price we can't match before we reach registry breadth.
  • Buyers won't switch off DIY/CPA absent a crisis, making CAC and sales cycles unworkable.
  • Ghost-asset savings disappoint, gutting the success-fee economics and the ROI pitch.
  • A self-caused missed deadline early on craters trust in a reference-driven market.
  • Licensing/contingency rules in target states prove costlier or more restrictive than modeled, slowing multi-state expansion.

27 90-day validation & launch plan

WeeksFocusActions & evidence gaps to close
1–2Buyer discovery15–20 interviews with tax/fixed-asset leaders at multi-state operators; confirm dissatisfaction, switching triggers, pricing tolerance (closes the biggest Unverified gaps).
2–4Legal & licensingCounsel review of beachhead-state (TX + 1–2) consultant licensing, agent-of-record, and contingency-fee rules; finalize engagement-letter templates.
3–6Engine MVPBuild ingest + normalization + beachhead jurisdiction registry + classification + ghost-asset flag + form-fill for the most common county forms.
5–8Free-scan wedgeRun ghost-asset/exposure scans on 5–10 prospects' real ledgers; measure actual ghost-asset % and dollar exposure (closes the recovery-hook gap).
6–10Pilot recruitmentSign 5–8 pilot accounts (cap enforced); instrument minutes/return, automation rate, consultant touch from day one.
8–12Pricing testTest per-return + success-fee vs. fixed-fee on pilots; validate margin trajectory.
10–13Compliance & QA hardeningStand up deadline-integrity reconciliation, audit-packet generation, second-eyes review; document SOPs.
OngoingKill criteriaStop if minutes/return don't fall across cycles, ghost-asset savings can't fund fees, consultant touch can't drop below ~100% without quality loss, or no pilot converts to program.

28 Sources

The ~38/36/14 state counts, the mandatory annual self-reported rendition with TX Apr-15 deadline (May-15 extension), the 10%/50% penalties and arbitrary-value treatment for non-filers, the 15–30% ghost-asset prevalence and its perpetual-overpayment effect, the TX TDLR property-tax-consultant licensing (and CPA/attorney exemption), the Avalara AvaMPT August-2025 launch, and the incumbent landscape are Verified via the sources above (retrieved 2026-06-30). Per-return fees, COGS, gross margin, revenue/FTE, automation rates, review-minute compression, ghost-asset savings magnitude, switching willingness, and bottom-up TAM are author estimates labeled Inferred or Unverified in §7/§15 and must be validated in pilots before decisive use. This is a hard-to-fool blueprint, not a guarantee.