DutyShield — The AI-Native AD/CVD Importer Compliance, Scope-Determination & Evasion-Defense Engine

Final decision: Blueprint

This run screened five candidates against the evidence threshold. One (FinCEN Residential Real Estate reporting) was killed on the spot when research showed a federal court vacated the rule on 19 March 2026, removing the compliance obligation entirely [S23]. The winner is DutyShield: an AI-native, per-unit-priced managed service that tells an importer, with an audit trail, whether each product it brings in is caught by an AD/CVD order, at what rate, and what it must do to stay out of a doubled-duty or evasion finding — then monitors that answer as orders, rates, and reviews change.

The timing is exceptional. The U.S. maintains hundreds of active AD/CVD orders — reported at 742 as of May 2025 and 600+ into early 2026, with 262 against China alone and dozens of new investigations underway [S3][S6]. CBP uncovered more than $400M in AD/CVD duty evasion in just the first eight months of 2025 and has made AD/CVD a Priority Trade Issue [S1][S3]. Combined duty rates on a single product can exceed 200% — a regular duty plus a Section 301 tariff plus a 200%+ AD margin can stack to 230% of value [S3][S17]. Under EAPA (19 CFR Part 165), any competitor or domestic producer can file an evasion allegation, and CBP can impose interim cash-deposit measures within 90 days [S7][S8]. Miss a non-reimbursement certificate and CBP presumes reimbursement and doubles the duties [S1][S18]. This is catastrophic, document-heavy, rules-bounded exposure that importers already pay $500–$600+/hr trade attorneys to manage [S21] — the textbook AI-native done-for-you service.

It clears every gate, with one honestly-handled caveat: a well-funded incumbent (Altana, $1B valuation, CBP contract) sells AI trade-compliance software the customer operates [S12][S13][S14]. DutyShield is deliberately the opposite — a managed outcome behind a licensed broker and attorney, not a screen the importer has to drive (see Anti-duplication and Anti-commoditization).

Executive summary

AD/CVD is the highest-consequence, least-forgiving corner of U.S. import compliance. An order can add a triple-digit-percent duty to a shipment; whether a given product is "in scope" is a genuinely hard legal-technical question decided by Commerce scope rulings and CBP entry review; rates are only estimated cash deposits, retroactively trued-up 12–18 months later through Commerce administrative reviews, so a company can owe (or be owed) large sums long after the sale [S1][S6]. Get any of it wrong — wrong scope call, missing certification, transshipment through a third country, an omitted non-reimbursement statement — and the importer of record faces doubled duties, interest, EAPA evasion findings with interim cash-deposit demands, civil penalties, and potential criminal referral [S1][S2][S7]. This is the outcome importers desperately want handled and already pay specialist trade counsel to handle at premium hourly rates [S21].

It is an ideal AI-native workflow: bounded, evidence-heavy, and recurring. The determination logic is rule- and document-driven — match a product's physical characteristics and HTS to the language of active orders and prior scope rulings; check country of origin and supplier against known transshipment and circumvention patterns; verify certification and non-reimbursement obligations; and quantify retroactive exposure across open review periods. Our engine ingests the importer's product specs, bills of materials, mill certs, supplier declarations, commercial invoices, and ACE entry data; uses OCR + LLM extraction and retrieval over the corpus of AD/CVD orders and Commerce scope rulings to produce a structured, cited scope-and-rate determination; runs deterministic checks (rate math, deadlines, certification completeness, review-period exposure); and routes every genuine judgment call and every filing to a licensed customs broker and, where the matter is legal, a trade attorney. We deliver the determination and the audit file; we file scope-ruling requests and EAPA responses through counsel; and we monitor each order-lane so the answer stays current as Commerce issues new orders, reviews, and rate changes [S4][S5][S9]. Because we are paid per determination, per monitored lane, and per defense package — plus a lawful contingency on duties actually recovered through refunds and protests — revenue recurs and incentives align. As frontier models improve, scope reasoning, order-corpus retrieval, and entry reconciliation get cheaper and sharper; first-pass accuracy rises and cost-per-unit falls, so margin compounds toward the mid-to-high 50s–60s. Importer counts and per-account revenue below are illustrative models, labeled accordingly.

Thesis

A large and growing population of U.S. importers is exposed to hundreds of active AD/CVD orders whose rates can exceed 200%, whose scope is genuinely ambiguous, and whose enforcement (EAPA, self-initiated CBP investigations, aggressive administrative reviews) is intensifying in the 2025–26 tariff environment. Importers already pay specialist trade attorneys and customs brokers premium fees to manage this, proving willingness-to-pay, but that help is labor-bound, episodic, and priced for large corporations — leaving the mid-market both most exposed and least served. An AI-native operation can profitably deliver the outcome — a defensible, audit-ready, continuously-monitored AD/CVD determination for every product-and-supplier lane — by collapsing the manual cost of product-to-order matching, scope-ruling retrieval, entry-data reconciliation, and evidence assembly, while keeping a licensed broker at the sign-off chokepoint and a trade attorney on every Commerce/CBP filing. We sell "you will never be blindsided by a retroactive AD/CVD bill or an evasion finding, and you have the file to prove your position" — an outcome, not software — and it grows stickier every time the importer adds a product or supplier and every time Commerce changes a rate.

Discovery rationale

The output root holds 165 prior blueprints, heavily concentrated in U.S. regulatory filings, trade compliance, and money-recovery plays. A duplicate scan found an adjacent trade cluster — hts-classification-duty-exposure-engine, duty-drawback-recovery-engine, ftz-compliance-filing-duty-recovery-engine, export-classification-licensing-engine, ocean-dd-dispute-recovery-engine, uflpa-forced-labor-traceability-detention-response-engine, fsvp-compliance-verification-engine, cbam-compliance-filing-engine, and eudr-deforestation-due-diligence-engine — but none covers antidumping/countervailing duty scope determination, certification, or EAPA evasion defense. AD/CVD is a distinct legal regime (Commerce + ITC + CBP) with its own scope-ruling process, certification rules, retroactive-review mechanics, and evasion-enforcement statute. Discovery targeted regulated, recurring, document-heavy problems where an ambiguous rule creates catastrophic downside and buyers already outsource to premium counsel. AD/CVD cleared the threshold most cleanly: hundreds of active orders, triple-digit rates, intensifying enforcement, verified willingness-to-pay, and a clean managed-service wedge that is clearly differentiated both from the covered trade engines and from Altana's customer-operated software.

Candidate comparison

CODE validation

Rubric scorecard — the six gates

Target buyer

Jobs-to-be-Done

• Functional: "Tell me, with proof, whether this product from this supplier is subject to an AD/CVD order, at what deposit rate, and what I must file — before I import, and again whenever the rules change."
• Functional: "Quantify and shrink my retroactive exposure across open review periods, and recover deposits I over-paid."
• Functional: "If CBP or a competitor accuses me of evasion, build the defense file fast."
• Emotional: "I don't want to lie awake over a seven-figure surprise duty bill I can't explain to the board."
• Social: "When the auditor, the board, or an acquirer asks, I can hand over a clean, defensible AD/CVD file instantly."

The painful problem

AD/CVD combines four uniquely nasty properties. (1) Catastrophic magnitude — rates can exceed 200% and stack with Section 232/301 tariffs to 230%+ of value [S3]. (2) Genuine ambiguity — whether a product is "in scope" turns on order language and a body of Commerce scope rulings; Commerce will initiate a scope inquiry, and if it does not act within 31 days the application is automatically initiated [S9]. (3) Retroactivity — deposits are estimates trued-up 12–18 months later in administrative reviews; a higher final rate means additional duties plus interest, long after the goods sold [S1][S6]. (4) Enforcement teeth — under EAPA (19 CFR Part 165) any interested party can allege evasion and CBP can impose interim cash-deposit measures within 90 days; a missing non-reimbursement certificate triggers a presumption of reimbursement and doubled duties; evasion can bring civil penalties and criminal referral [S1][S7][S18]. CBP found $400M+ in evasion in eight months of 2025 [S1][S3]. The mid-market importer is exposed to all of this but cannot afford $600/hr counsel on every SKU — so the liability sits unmanaged.

The outcome we sell

Concretely, per lane the client receives: a scope-and-rate determination memo (in-scope? which order? deposit rate? certification/non-reimbursement obligations?), the cited evidence (order language, matching scope rulings, mill certs, origin analysis), a retroactive-exposure estimate across open review periods, a monitoring subscription that re-fires the determination when Commerce changes anything, and — when needed — a scope-ruling request or EAPA-defense package prepared through counsel.

First one-feature MVP wedge

Evidence summary

The core drivers are Verified from primary and authoritative secondary sources: hundreds of active AD/CVD orders and ongoing new investigations (Commerce/CBP/Federal Register) [S3][S4][S5][S6]; the scope-ruling process and 31-day auto-initiation (Commerce scope-ruling guide) [S9]; EAPA mechanics and interim measures (19 CFR Part 165, CBP) [S7][S8]; the doubled-duty non-reimbursement rule and certification requirements (CBP FAQ, Tradewin) [S1][S18]; retroactive administrative-review truing (CBP/Commerce) [S1][S6]; $400M+ evasion found in 8 months of 2025 [S1][S3]; 200%+ / 230% stacked rates [S3][S17]; premium trade-counsel rates [S21]; and the incumbent-software landscape (Altana funding and CBP contract) [S12][S13][S14]. Market-size and per-account revenue figures are Inferred models, labeled as such. The order-count figures (742 / 600+ / 262) come from aggregator and law-firm summaries citing Commerce; treated as Verified-with-range given multiple concurring sources but flagged as point-in-time.

Claim table

*Verified-with-range: order counts are point-in-time and vary slightly by source; the order of magnitude (hundreds, growing) is consistently supported.

Source-claim matrix

Market & demand evidence

The U.S. maintains hundreds of active AD/CVD orders across steel & aluminum (150+ orders), chemicals, solar, wood/furniture/cabinets, mattresses, tires, quartz, and fasteners, with new investigations opening continuously and annual administrative reviews churning rates for named companies [S3][S4][S5]. Section 232 metals tariffs were expanded and recalibrated in 2025 and Section 301 duties layer on top, pushing more importers to re-source and re-open scope questions [S15][S16]. CBP elevated AD/CVD to a Priority Trade Issue, found $400M+ in evasion in 8 months of 2025, and adopted AI (Altana) for enforcement targeting in late 2025 — a clear signal that the enforcement side is getting sharper, which raises importers' need for a defensible position [S1][S13]. A mature ecosystem of trade-law firms and customs brokers already sells AD/CVD scope, certification, and EAPA-defense work, confirming a real, funded budget line [S19].

Active buyer conversations

• Trade-law firm client alerts and FAQs answering "is my product covered," "what is a scope ruling," and "I received an EAPA allegation" — a continuous stream indicating live buyer questions [S1][S17][S19].
• Importer and customs-broker communities (e.g., trade-compliance groups on LinkedIn, the ICPA community, freight forwarder forums) repeatedly surfacing scope, transshipment, and certification questions.
• CBP's own public FAQ and Commerce's scope-ruling guide exist precisely because importers keep asking — the government publishes buyer-facing guidance because the demand for answers is constant [S1][S9].
• Section 232/301 re-sourcing content and webinars in 2025–26 driving importers to re-examine AD/CVD exposure on new suppliers [S15].

Competitive landscape

Competitor & budget validation

Budget already exists and is substantial: importers pay trade attorneys premium hourly fees and retain customs brokers, and a whole trade-law sub-industry markets AD/CVD services [S19][S21]. Enterprise trade-compliance software is well-funded (Altana at a $1B valuation) — proof of category budget, but aimed at large enterprises and sold as software the customer operates [S14]. The current alternatives are insufficient for the mid-market: law firms are too expensive per-SKU and episodic; brokers lack AD/CVD depth and are conflicted; enterprise SaaS requires in-house expertise. DutyShield wins by redirecting the existing legal/brokerage spend into a productized, per-unit managed outcome that is cheaper than hourly counsel, deeper than a broker's guess, and requires no software operation — while keeping the licensed sign-off buyers need to actually transfer the liability. It is not a clone of Altana (software) or of a law firm (bespoke hourly); it is a new operating model between them.

Pricing evidence & proposed pricing

Anchor: specialist trade counsel at $500–$600+/hr, where a single scope-ruling engagement or EAPA defense easily runs tens of thousands of dollars [S19][S21]. We price the outcome, well below hourly counsel, with recurring monitoring.

Never hourly. Recovery contingency is applied only to money actually returned to the client, avoiding any incentive to under-declare duties owed.

Regulatory & compliance considerations

Three regimes govern the work. Commerce (Enforcement & Compliance) issues and administers AD/CVD orders, conducts scope inquiries (auto-initiating if it doesn't act in 31 days), runs administrative reviews that retroactively set final rates, and codified importer/exporter certification requirements in recent rulemaking [S6][S9][S10][S11]. CBP collects deposits, liquidates entries, applies the doubled-duty presumption when the non-reimbursement certificate is missing, and enforces EAPA (19 CFR Part 165) — where any interested party can allege evasion and CBP can impose interim cash-deposit measures within 90 days [S1][S7][S8][S18]. Circumvention/evasion law (transshipment, misclassification, undervaluation) exposes importers to penalties and criminal referral [S20]. The importer of record bears ultimate legal responsibility; our role is to produce defensible analysis and evidence and to file, through appropriately licensed people, on the client's behalf.

Licensing boundary

Disclaimers, engagement letters defining the importer-of-record's retained responsibility, POAs, consent for account access, and full audit logs are standard on every engagement. We do not provide unlicensed legal advice or transact customs business without a broker.

AI-native advantage

AI changes the economics, not just the vibe. The determination is fundamentally a retrieval-and-synthesis problem over a large, structured corpus (active orders, hundreds of scope rulings, HTS, origin rules) applied to messy inputs (specs, BOMs, mill certs, invoices, ACE data). Manually, an analyst or attorney reads order language and hunts prior rulings for hours per SKU. Our engine does the extraction, matching, retrieval, and drafting in minutes, leaving the human to judge and sign. That collapses cost-per-determination by an order of magnitude and lets us profitably serve the mid-market that counsel prices out — while producing a more consistent, fully-cited file than a hurried human read. What must never be fully automated: the final scope call in ambiguous cases, any filing with Commerce/CBP, and any legal position — all reserved for the licensed chokepoint.

Internal AI engine architecture

AI-vs-human operations pipeline

Blue = AI · Green = deterministic rules · Amber = licensed human chokepoint. Humans concentrate on judgment and filings; everything else automates.

Dynasty translation layer

Anti-duplication analysis

What exists: trade-law firms (bespoke hourly), customs brokers (entry filing), enterprise trade-compliance SaaS (Altana, Descartes, SAP GTS, Thomson Reuters), and — in our own catalog — HTS classification, duty drawback, FTZ, export classification, UFLPA, FSVP, CBAM, and EUDR engines. Why this is not a copy: none of them productizes AD/CVD scope-and-rate determination-plus-monitoring as a per-unit managed outcome for the mid-market with a licensed sign-off. AD/CVD is a legally distinct regime (Commerce scope rulings, retroactive reviews, EAPA) not addressed by the covered HTS/drawback/FTZ engines — HTS classification answers "what tariff code," AD/CVD answers "is this caught by a punitive order and at what retroactive rate." The narrow wedge: one product/supplier lane → one cited, broker-signed determination. Under-served segment: mid-market importers priced out of $600/hr counsel and lacking in-house trade expertise. Unsolved pain: continuous, defensible, per-SKU exposure management with an audit trail. Our differentiation is the operating model (managed outcome + licensed chokepoint), the proprietary determination corpus, and the liability transfer — not raw AI.

Anti-commoditization analysis

Sam Altman test: better models make scope reasoning, order-corpus retrieval, and entry reconciliation cheaper and sharper — the engine strengthens, not obsoletes. If general models let importers self-serve some determinations: we still win because (1) the importer needs a licensed broker/attorney sign-off to actually rely on and transfer the liability — a general chatbot cannot sign a customs deliverable or file with Commerce; (2) our proprietary, continuously-updated corpus of orders, scope rulings, and actual client outcomes outperforms a cold model; (3) monitoring and deadline management is an operational service, not a one-shot answer; and (4) buyers demand accountability and audit trails they can show CBP, an auditor, or an acquirer. The commoditizable part (drafting) is exactly the part we automate; the defensible part (licensed judgment, corpus, liability, operations) is what we sell.

Service delivery workflow

1. Onboard: engagement letter defining importer-of-record responsibility; POA and account-access consent; lane inventory.
2. Intake: collect specs, BOMs, mill certs, supplier declarations, invoices, HTS, ACE data per lane.
3. Determination: AI drafts cited scope-and-rate memo; rules layer computes rate, certification obligations, and exposure.
4. Review: licensed broker reviews/sign; attorney escalation for ambiguity; second reviewer over exposure threshold.
5. Deliver: memo + evidence file + exposure dashboard; recommended actions.
6. Monitor: subscription watches Federal Register/Commerce; re-fires determination on any change; tracks review and certification deadlines.
7. Defend / recover: attorney-led scope-ruling requests, EAPA defense, and protest/refund recovery as triggered.
8. Renew: annual retainer for exposure/certification file; expand lanes.

Operations as product

The operation is the product. We eliminate variance with: SOPs for each determination type; structured intake checklists and required-evidence lists per sector; automated completeness checks that reject incomplete lanes before analysis; an exception queue for ambiguous scope; reviewer-assignment logic by sector and exposure; confidence scoring; full audit trails and version control on every memo; a gold-standard library of past determinations and actual CBP/Commerce outcomes; red-team checks on high-exposure lanes; standardized client-ready memo templates; and root-cause + postmortem loops on any determination later challenged or overturned. Each correction becomes an SOP, rule, prompt, retrieval source, or QA check — so the system gets more consistent and cheaper per unit over time.

No-holes quality engine

• Completeness gate: a lane cannot be analyzed until required evidence (spec, origin proof, HTS, supplier data) is present.
• Dual-key on high exposure: any lane above an exposure threshold gets a second licensed reviewer.
• Citation enforcement: every scope conclusion must cite order language and/or a scope ruling; uncited conclusions are blocked.
• Confidence routing: low-confidence lanes auto-route to attorney escalation, not to the client.
• Outcome reconciliation: every actual CBP/Commerce result is compared to our determination; misses trigger a postmortem and a gold-example update.
• Change-monitoring SLA: new orders/reviews affecting a monitored lane trigger client notice within a defined window.

What the human expert actually does

Minimum viable offer

"AD/CVD Lane Determination — $2,500, delivered in 3 business days, broker-signed." You send us one product-and-supplier lane; we tell you, with cited proof, whether it's caught by an AD/CVD order, at what deposit rate, and what you must file — plus your retroactive-exposure flag. First determination credited toward a monitoring plan if you subscribe. This is the whole wedge: one lane in, one defensible answer out.

Fulfillment process (first 3 customers, manually)

1. Founder (trade attorney or ex-broker) closes 3 pilot importers via warm GovCon/trade network.
2. Intake by shared folder; AI drafting via off-the-shelf LLM + a hand-built index of relevant orders and scope rulings.
3. Founder/broker reviews and signs each determination personally.
4. Deliver memo + evidence file as PDF; capture every correction.
5. Stand up a lightweight monitoring watch (Federal Register alerts) manually before automating.
6. Convert each determination into an SOP and a gold example; harden intake before adding customers.

Tools & systems

Day one: secure portal/SFTP, document AI/OCR, a frontier LLM with a curated index of active orders + Commerce scope rulings, a spreadsheet/rules layer for rate and deadline math, case-management/CRM (e.g., a simple pipeline tool), e-sign, and Federal Register/ACCESS email alerts. Later: automated corpus refresh from the Federal Register and Commerce feeds, ACE/ABI data integration, an exposure dashboard, and a determinations database that becomes a proprietary asset. Favor available tools before custom software; build the corpus and rules engine as the durable IP.

Human-in-the-loop quality control

Nothing reaches the client or a government filing on AI output alone. The broker signs every determination; the attorney owns every Commerce/CBP filing and every ambiguous scope call; high-exposure lanes get a second reviewer; low-confidence lanes escalate automatically. Every determination carries a citation trail, and every actual government outcome is reconciled against our call to catch and learn from misses.

Nonlinear scaling & unit economics

All figures below are illustrative models, not guarantees.

Revenue scales faster than headcount because the AI absorbs the linear extraction/retrieval/drafting cost while the licensed human is concentrated at sign-off and filings. Margin expands as review minutes fall and the corpus and gold-example library reduce escalation. CAC payback target <6 months on the determination+monitoring bundle. Funnel assumptions (models): lead-magnet (free exposure scan) → determination conversion ~15%; determination → monitoring subscription ~40%; monitoring annual retention ~85%.

Distribution proof table

Sales & outreach plan

Lead with a diagnosis, never a demo. The three-layer motion: (1) expert-led content teaching importers how AD/CVD scope, certification, and EAPA actually work; (2) warm GTM to the founder's trade/GovCon network and to customs brokers who can white-label; (3) targeted outbound using public import data to identify importers of exposed goods and send a personalized exposure memo ("Your recent entries of [product] from [country] may be caught by [order] — here's what we'd check"). Every path funnels to the free exposure scan → paid determination → monitoring subscription.

Founder-led content plan

Publish weekly from a licensed trade-expert voice: scope-ruling teardown of a real recent Commerce ruling; "what a CF-28 really means"; "how administrative reviews retroactively wreck your margin"; "the non-reimbursement certificate that doubles your duties"; "EAPA allegation — the first 72 hours"; sector deep-dives (steel, solar, furniture); and myth-busting ("my broker cleared it, so I'm fine"). Repurpose the highest-performing posts into paid-ad tests and answer-engine content.

First 30 days of content

10 educational posts

1. Is your product AD/CVD-covered? The 5-question scope test.
2. The $400M signal: what CBP's 2025 evasion haul means for you [S1].
3. How a 5% duty becomes 230% (232 + 301 + AD stacking) [S3].
4. The non-reimbursement certificate that doubles your duties [S18].
5. Scope rulings 101: the 31-day auto-initiation trap [S9].
6. Retroactive reviews: why "cleared at entry" isn't final [S1].
7. EAPA: how a competitor can put you under interim measures in 90 days [S7].
8. Transshipment red flags CBP's AI now catches [S13].
9. Section 232 changed — re-check your suppliers [S15].
10. "My broker said it's fine": why that's not a defense.

3 diagnostic teardown formats

1. Live scope teardown of a real product on video.
2. Anonymized "exposure memo" walkthrough.
3. Before/after: an importer's unmanaged vs. managed AD/CVD file.

2 lead-magnet angles

1. Free AD/CVD Exposure Scan (upload your product list).
2. Sector cheat-sheet: active orders in steel / solar / furniture.

1 webinar

"Surviving the next administrative review: a live AD/CVD exposure clinic."

1 outbound diagnosis template

"We reviewed public entry data for [importer]. Your imports of [product] from [country] may fall under [order] at [rate]. Here's a free 1-page exposure read."

Lead magnet & waitlist plan

Lead magnet: the Free AD/CVD Exposure Scan — the importer uploads a product/supplier list; the engine returns a one-page read flagging which lines are likely caught by an active order and roughly at what rate, with a clear "these need a full determination" list. It captures the exact pain signal (which lanes are exposed), builds trust by being specific and cited, and hands the operator a warm, qualified list. Waitlist: for the monitoring product, gated by exposure-scan completion. Sales-ready = has ≥1 flagged high-exposure lane and an active importer-of-record account.

Warm GTM plan

Work the founder's trade-law/broker/GovCon network and every exposure-scan and webinar participant. Offer a free full determination on their single scariest lane to prove value, then convert to monitoring. Prioritize customs brokers for white-label partnerships — they bring importer books and want AD/CVD depth without hiring counsel.

Targeted outbound plan

Use public import data (ACE summaries, bill-of-lading/manifest datasets) to identify importers of AD/CVD-exposed goods, then send a personalized, diagnosis-first exposure memo — not a demo ask. Sequence: exposure memo → offer a free determination on one lane → pilot. Personalize on the actual product, country, and likely order.

Answer-engine / search visibility plan

Importers increasingly ask ChatGPT/Perplexity/Google "is [product] from [country] subject to antidumping duties?" Publish structured, citable explainers on scope rulings, EAPA, certifications, and sector orders so DutyShield is the cited authority in AI answers. Maintain a public, regularly-updated "active orders by sector" reference and scope-ruling explainers optimized for LLM retrieval, each ending in the free exposure scan.

Pilot design & early-demand-trap mitigation

First cohort: 5 mid-market importers in steel/aluminum, solar, or wood products. Cap: hard cap at 5 until intake and QA are hardened. Incentive: free first determination + discounted 6-month monitoring in exchange for weekly feedback and permission to use anonymized outcomes. Feedback: weekly 30-minute review of every determination and correction. Product feedback vs custom work: a recurring determination pattern or rule gap = product; a one-off bespoke legal strategy = custom (attorney-billed, not scaled). Corrections become SOPs, rules, prompts, retrieval sources, and gold examples. Avoid the trap of absorbing unlimited bespoke legal work under a productized banner.

Early-access feedback flywheel

Every broker/attorney correction and every actual CBP/Commerce outcome is captured, labeled, and folded back into the engine as a gold example, a prompt/rule update, or a new retrieval source. First-pass broker-agreement rate and overturn rate are the north-star quality metrics; both should climb pilot-over-pilot. Before expanding beyond 5, the AI draft must reach ≥85% broker agreement on covered sectors.

Build-before-scale checkpoints

• After 5 importers: harden intake, required-evidence lists, and QA gates; codify the completeness gate.
• After 10: harden SOPs, exception queues, reviewer checklists, and memo templates; automate corpus refresh and monitoring.
• After 20: pause new pilots until COGS, rework, escalation, cycle time, and overturn rate are measured and within target. Acceptable temporary workaround: manual Federal Register watch. Unscalable signal: every determination needing bespoke attorney judgment (means the wedge is too broad and must be narrowed by sector).

7-day launch plan

1. Stand up the exposure-scan landing page + waitlist.
2. Build the initial index of active orders + scope rulings for 3 target sectors.
3. Draft engagement letter, POA, and licensing-boundary disclaimers with counsel.
4. Publish 3 cornerstone posts + the outbound exposure-memo template.
5. Line up a licensed customs broker and a partner trade attorney.
6. Hand-run 3 sample determinations to validate the AI draft → broker sign flow.
7. Book 5 pilot conversations from the warm network.

30-day launch plan

1. Sign 5 pilot importers; deliver first determinations.
2. Ship the free exposure scan; drive the first 100 scans via content + outbound.
3. Run the "administrative review survival" webinar.
4. Sign the first customs-broker white-label partner.
5. Codify SOPs and gold examples from pilot corrections; automate Federal Register monitoring.

90-day launch plan

1. Convert ≥40% of determinations into monitoring subscriptions.
2. Reach ≥85% broker agreement on AI drafts in covered sectors.
3. Deliver the first scope-ruling request and/or EAPA-defense package through counsel.
4. Measure COGS, rework, escalation, cycle time, overturn rate; hit build-before-scale gates.
5. Expand to a 4th sector and a second broker partner; formalize the recovery-contingency offer with legal review.

Metrics & KPIs

• Quality: broker first-pass agreement rate; determinations overturned by CBP/Commerce (<2% target); rework rate.
• Speed: cycle time/determination; monitoring change-notice SLA adherence.
• Economics: gross margin; revenue/FTE; COGS/determination; CAC payback.
• Funnel: scans → determinations (~15%); determinations → monitoring (~40%); monitoring retention (~85%).
• Growth: active monitored lanes; broker partners; recovery dollars returned to clients.

Risks & mitigations (summary)

The material risks are: a wrong determination causing client harm and liability; unauthorized customs business or UPL; incumbent (Altana) or law-firm competition; regulatory change; corpus staleness; contingency-fee legality; data security; talent scarcity (licensed brokers/attorneys); tariff-regime volatility; and client concentration. Each is addressed in the register below.

Exhaustive risk register

What could kill this

The two real killers: (1) a determination we signed turns out badly wrong and the client suffers doubled duties or an evasion finding — an existential trust and liability event, which is why the licensed sign-off, dual-key, citation enforcement, E&O insurance, and importer-of-record responsibility framing are non-negotiable; and (2) a well-funded incumbent decides to productize a mid-market managed AD/CVD service with licensed sign-off — mitigated by moving fast on the broker-partner channel and the proprietary determination/outcome corpus. A distant third is a wholesale political dismantling of the trade-remedy regime, which is unlikely given bipartisan support for AD/CVD and hundreds of durable active orders.

Go/no-go reasoning

Every evidence-threshold test is met: a clearly identified buyer (mid-market importer CFO/compliance lead); a painful, specific, catastrophic problem (AD/CVD exposure); verified evidence the problem exists and is enforced ($400M evasion, hundreds of orders, EAPA, doubled duties); verified existing spend (premium trade counsel and brokers); active demand (constant scope/certification/EAPA questions); competitor and budget validation (law firms + funded SaaS); a credible reason to win (managed outcome + licensed sign-off + mid-market focus); a narrow MVP wedge (one lane → one determination); a service-first delivery path (manual first 3, no big platform needed); no unresolved fatal blocker (licensing handled via broker + attorney); a credible path to 55–62% gross margin; and a believable distribution path (brokers, content/AEO, outbound on import data). The lone caveat — Altana-style software competition — is exactly why the model is a licensed managed service, not a tool.

Final recommendation

Build it. DutyShield sells the one thing importers in AD/CVD-exposed sectors most want handled and most fear getting wrong: a defensible, continuously-monitored, per-lane duty-exposure determination, signed by a licensed expert, at mid-market prices. It clears all six gates (28/30), passes the evidence threshold, is cleanly differentiated from every covered trade engine and from Altana's customer-operated software, prices strictly per-unit with a lawful recovery contingency, and strengthens as frontier models improve. Start with the $2,500 single-lane determination for steel/solar/wood importers, prove ≥85% broker agreement across a 5-importer pilot, then expand into monitoring, recovery, and broker white-label.

Sources

1. [S1] CBP — Antidumping and Countervailing Duties (AD/CVD) FAQ — cbp.gov
2. [S2] Camtom — AD/CVD Evasion: EAPA Investigations in 2026 — camtomx.com
3. [S3] Camtom — Antidumping & Countervailing Duties Explained for US Importers (2026) — camtomx.com
4. [S4] Federal Register — Opportunity To Request Administrative Review (2026-01-02) — federalregister.gov
5. [S5] Federal Register — Initiation of AD/CVD Administrative Reviews (2026-01-05) — federalregister.gov
6. [S6] U.S. Commerce (trade.gov) — U.S. Antidumping and Countervailing Duties — trade.gov
7. [S7] eCFR — 19 CFR Part 165 (Investigation of Claims of Evasion / EAPA) — ecfr.gov
8. [S8] CBP — Enforce and Protect Act (EAPA) — cbp.gov
9. [S9] U.S. Commerce — Scope Ruling Application Guide — access.trade.gov (PDF)
10. [S10] Wiley — Commerce Publishes Final Rule Amending AD/CVD Procedures — wiley.law
11. [S11] ArentFox Schiff — New Commerce Regulations Governing AD/CVD Now in Effect — afslaw.com
12. [S12] Altana — The AI-Powered Network for Trusted Trade — altana.ai
13. [S13] Axios — Exclusive: CBP taps Altana's AI for trade enforcement (2025-10-31) — axios.com
14. [S14] Crunchbase — Altana Company Profile & Funding — crunchbase.com
15. [S15] BDO — Section 232 Metals Tariffs Expanded and Recalibrated — bdo.com
16. [S16] CBP — Section 232 Tariffs on Steel and Aluminum FAQ — cbp.gov
17. [S17] Harris Sliwoski — U.S. Customs Compliance FAQ (2025) — harris-sliwoski.com
18. [S18] Tradewin — AD/CVD Certification Requirements — blog.tradewin.net
19. [S19] Great Lakes Customs Law — AD/CVD Duties and Scope Rulings — greatlakescustomslaw.com
20. [S20] U.S. Court of International Trade — Circumvention, Evasion, and Enforcement of AD/CVD Orders — cit.uscourts.gov (PDF)
21. [S21] Clio — Compare Average Lawyer Hourly Rate by State (2026) — clio.com
22. [S22] FreightFigures — Customs Bonds Explained (2026) — freightfigures.com
23. [S23] Foley & Lardner — Federal Court Vacates FinCEN Residential Real Estate Reporting Rule (2026-03) — foley.com