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

CBAM Declarant Desk — Done-for-You EU Carbon Border Declaration Engine

A done-for-you service for EU importers of CBAM goods (above the 50-tonne threshold) that delivers one coupled outcome each year: a complete, verifier-ready, registry-filed annual CBAM declaration with the lowest defensible certificate liability — supplier installation-level emissions data collected and validated for you, embedded emissions calculated under the EU methodology, punitive default values displaced by verified actual data wherever it pays, and quarterly certificate-holding forecasts kept current in between. AI is the internal production line; a carbon-accounting specialist is the customer-facing expert, and the statutory third-party verification stays with an independent EU-accredited verifier.

Run: 2026-06-30 · Hour 11 · NN 01 Sector: EU trade & customs carbon compliance — Carbon Border Adjustment Mechanism Buyer: Trade-compliance / procurement / finance leaders at EU importers of CBAM goods Pricing: Per-declaration managed program + per-supplier-installation data unit (+ optional default-avoidance share) Outcome: Verifier-ready, registry-filed CBAM declaration with minimized certificate cost

01 Thesis

On 1 January 2026 the EU Carbon Border Adjustment Mechanism entered its definitive regime: importers of iron & steel, aluminium, cement, fertilisers, hydrogen and electricity no longer just file quarterly reports — they must hold authorised CBAM declarant status, submit one annual CBAM declaration (first due 30 September 2027 for 2026 imports), and buy and surrender CBAM certificates for every tonne of embedded CO2 (Verified, EC + DEHSt + Reg (EU) 2023/956). The first published certificate price was €75.36/tCO2e for Q1 2026 (Verified). Under-surrender carries a €100/tCO2e penalty that does not cancel the surrender obligation; importing without declarant status carries €300–500/tCO2e (Verified, Art. 26 as amended by Reg (EU) 2025/2083).

The cost of the declaration now turns on one brutal operational fact: supplier installation-level emissions data is the single biggest bottleneck (Verified, CarbonChain/Zevero/Coolset). If a non-EU supplier will not supply verified actual emissions, the importer falls back to EU default values that carry a deliberate mark-up — +10% in 2026, +20% in 2027, +30% from 2028 (Verified, IR (EU) 2025/2621) — and that can sit 80–130% above a well-run installation's real intensity. Every tonne overstated is an unnecessary certificate bought at ~€75. We sell the outcome: run the supplier-data campaign, compute the embedded emissions under the EU rules, drive verified actual data wherever it pays, coordinate the accredited verifier, and hand over a registry-ready declaration plus quarterly certificate forecasts.

This is not a co-pilot. A trade-compliance manager importing structural steel from three countries does not want yet another dashboard that surfaces 40 suppliers for her to chase in five languages; she wants the data collected, the emissions calculated, the certificate bill minimized, and the declaration filed, by an expert who is accountable for it. The work is document-and-data synthesis against fixed EU methodology and CN-code mappings, so it decomposes into automatable steps with judgment concentrated at a few chokepoints — methodology/system-boundary selection and verifier coordination — the shape that lets revenue scale faster than headcount toward software-like margins.

02 Discovery rationale

This run scanned regulated, document-heavy, deadline-driven compliance terrain where work is commonly outsourced and decomposes cleanly: EU carbon/trade compliance, deforestation due diligence, product-carbon-footprint regimes, US chemical reporting, and mortgage/financial licensing. Five candidates were generated and three deep-validated (see §3). The CBAM Declarant Desk won on a rare combination: (a) a live, binding obligation that just switched on (definitive regime active since 1 Jan 2026) with hard money attached (€100/t and €300–500/t penalties, ~€75/t certificate cost); (b) a concentrated, identifiable buyer — the Omnibus 50-tonne threshold exempts ~90% of importers by number but keeps 99% of emissions, so spend funnels to a knowable set of larger importers (>4,100 authorised declarants already by January 2026); (c) a recovery hook — displacing punitive default values with verified actual data produces a provable euro saving on the very first cycle; and (d) an existing, validated budget already flowing to Big Four CBAM desks (€15k–40k/quarter), boutiques, and software-plus-service vendors — proving the category while leaving a done-for-you mid-market wedge open.

The decisive evidence was the asymmetry between how mechanical the work is (CN code → production route → precursor map → EU default vs. verified actual → embedded-emissions math → declaration) and how painful it is at scale (multilingual supplier chasing, fragmented data across tiers and geographies, tightening verifier capacity, real cash penalties). That asymmetry is exactly where an internal AI engine plus a thin specialist-review layer can deliver a done-for-you outcome at a margin that a consultant re-reading the same Turkish mill emissions declaration 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
CBAM Declarant Desk Trade-compliance / procurement / finance at EU importers of CBAM goods (esp. iron & steel) Verifier-ready, registry-filed annual CBAM declaration with minimized certificate liability 4.4 Strong SELECTED — live binding obligation, hard penalties, concentrated buyer, recovery hook, validated budget.
EUDR Due-Diligence Statement Desk EU operators/importers of soy, beef, palm, wood, cocoa, coffee, rubber Filed Due Diligence Statements (TRACES) so shipments clear customs 3.5 Verified, lower urgency Rejected — application delayed to 30 Dec 2026 (Reg (EU) 2025/2650); downstream operators no longer file separate DDS (smaller pool); geolocation/GeoJSON plot data leans toward physical traceability; rules still in “simplification review” (Apr 2026).
EU Battery Regulation Carbon-Footprint Declaration Desk Producers/importers of EV, LMT & industrial (>2 kWh) batteries Conformity-grade battery carbon-footprint declaration + passport-ready data 3.2 Inferred Rejected — buyer pool is narrow (cell/battery makers); requires full PEF life-cycle studies (specialist LCA); several obligations are contingent on delegated/implementing acts not yet in force; thin active-demand evidence vs. CBAM.
PFAS TSCA §8(a)(7) Reporting Desk (US) US manufacturers/importers of PFAS or PFAS-containing articles since 2011 Filed retrospective PFAS reports to EPA 3.0 Inferred Rejected — largely a one-time retrospective filing → weak recurrence/retention; submission-window timing has repeatedly shifted; one-shot revenue undermines the recurring-engine thesis.
NMLS Mortgage Call Report / License-Renewal Desk (US) State-licensed mortgage lenders/brokers & MLOs Filed quarterly MCRs + completed annual license renewals 2.9 Inferred Rejected — adjacent to the prior mortgage post-close QC blueprint; crowded compliance-vendor field; structured quarterly form = lower intelligence threshold; weaker regulation-as-moat differentiation.

04 Hard disqualifier check

#DisqualifierStatusNote
1Customer-facing co-pilot / SaaS, not done-for-youPassWe run the supplier-data campaign, compute emissions, coordinate verification and file; the importer hands over its import & supplier list and receives a finished declaration.
2Requires physical labor / field crews / site visitsPassPure data/document orchestration. The physical installation audit is the accredited verifier's job, not ours; we coordinate it.
3Primary pricing is hourly / cost-plusPassPer-declaration managed program fee + per-supplier-installation data unit; optional default-avoidance savings share. No hourly billing.
4Cannot plausibly reach 50%+ gross marginUnclearPlausible as supplier-outreach and extraction automate and the supplier/installation dataset is reused across importers; unproven — central pilot kill-metric (§15).
5Buyer cannot be identifiedPassAuthorised CBAM declarants & importers above 50 t — a registrable, concentrated set (>4,100 by Jan 2026); iron & steel dominates volume.
6Workflow cannot be decomposedPassImport list → supplier/installation map → data request → extraction → CN/route/precursor mapping → default-vs-actual → emissions calc → declaration draft → verifier coordination → file & reconcile certificates.
7Fully automates regulated judgment without licensed/expert reviewPassMethodology/system-boundary calls reviewed by a carbon-accounting specialist; statutory verification is performed by an independent EU-accredited verifier — never by us, never by the model.
8Substantially duplicative of a prior blueprintPassDistinct statute (Reg (EU) 2023/956), data (installation GHG of imported goods), trigger (CBAM declaration / customs border), deliverable (declaration + certificate reconciliation) vs. conflict-minerals, packaging-EPR, climate-disclosure, environmental-reporting (§20, manifest notes).
9Likely illegal / un-incorporable licensingPassWe are a preparer/orchestrator, not the accredited verifier and not (by default) the declarant of record; indirect-customs-representative rules handled explicitly (§21).
10Core demand unverified / not inferablePassBinding annual obligation + cash penalties + €75/t certificates + >12,000 authorisation applications = Verified demand.
11Frontier models commoditize rather than strengthenUnclearSoftware incumbents (CarbonChain, CBAMBOO, Dubrink) also benefit; moat must come from done-for-you accountability, a reusable supplier/installation dataset, and verifier coordination, not model access (§17).
12Cannot be tested with a small bounded pilotPassOne importer, one steel-import program, one declaration cycle, measured default-vs-actual savings — a clean bounded test.

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

05 Rubric scorecard

Low trust burden
4.4
Low task-level judgment
4.2
High intelligence threshold
4.3
Regulation as moat
4.3
No physical labor
4.6
Sam Altman test
3.9

Aggregate ≈ 4.3 / 5. Strongest on no-physical-labor and regulation-as-moat (binding EU law, cash penalties, accredited-verifier gate, customs enforcement). Weakest on the Sam Altman test — better models also help the customer-operated CBAM software tools; the durable edge is the done-for-you accountable outcome, the reusable verified supplier/installation dataset, and the verifier-coordination relationship, not raw model access.

06 Opportunity

€89bn
2024 EU imports covered by CBAM (500+ goods) V
~105 Mt
2024 CBAM goods volume; iron & steel ≈ 69% V
€100/t
Penalty for under-surrender (Art. 26(1)); €300–500/t unauthorised V
€75.36
First CBAM certificate price /tCO2e (Q1 2026) V
>4,100
Authorised CBAM declarants by early Jan 2026 (12,000+ applied) V
50 t
De minimis threshold: ~90% importers exempt, 99% emissions kept V
+10→30%
Default-value mark-up 2026→2028 (IR (EU) 2025/2621) V
30 Sep 2027
First annual CBAM declaration deadline (for 2026 imports) V

CBAM compliance is large, freshly mandatory, recurring, and quietly expensive. Above the 50-tonne line, an importer must register, collect installation-level emissions from non-EU suppliers (often across multiple tiers and languages), compute embedded emissions under EU methodology, and then pay ~€75 for each tonne it cannot evidence as lower. Two euro pools are on the table: the penalty/risk euros from late, missing, or unauthorised declarations, and the over-payment euros from relying on marked-up default values where verified actual data would be materially lower. The second is a perpetual, compounding leak — and the lever that lets a recurring engine prove euro savings on the very first cycle (Verified mechanism; per-account magnitude is account-specific, §7).

07 Evidence quality & source-claim matrix

ClaimLabelSource / basisConf.Business impact
Definitive regime live since 1 Jan 2026; quarterly reports replaced by one annual declaration; first due 30 Sep 2027 for 2026 importsVerifiedEC CBAM Q&A; EC “entered into force” news (Jan 2026); DEHSt; Reg (EU) 2023/956HighDefines the recurring, deadline-driven job we own.
Authorised CBAM declarant status mandatory; apply by 31 Mar 2026 to keep importing; goods stopped at border otherwiseVerifiedNet0; ReedSmith; Compliance & Risks; EC Q&AHighCreates the urgent trigger + customs-enforcement teeth.
50-tonne de minimis (Reg (EU) 2025/2083): ~90% of importers exempt by number, 99% of emissions retained; ~182,000 importers excluded; not for electricity/hydrogenVerifiedReg (EU) 2025/2083; ICAP; ReedSmith; Council ST-16973-2025HighConcentrates the buyer pool to larger, knowable importers.
>12,000 authorisation applications by 7 Jan 2026; >4,100 obtained declarant status; iron & steel = 98% of first-window declared volumeVerifiedEC Taxation & Customs Union news (14 Jan 2026)HighDirect, current demand signal & beachhead sector.
2024 CBAM-covered imports ≈ €89bn / ~105 Mt; iron & steel ≈ 69% of volume; fully phased-in carbon cost ≈ €12bnVerifiedCO2 IQ (Eurostat-based)Med-HighSizes the aggregate spend the obligation creates.
Penalties: €100/tCO2e under-surrender (Art. 26(1)); €300–500/t unauthorised (Art. 26(2)); penalty does not cancel surrender obligationVerifiedCBAM Guide (penalties/omnibus); DEHSt sanctioning; conformi.eu; Reg (EU) 2025/2083HighQuantifies the downside the customer pays us to prevent.
First certificate price €75.36/tCO2e (Q1 2026); sales open 1 Feb 2027; 50% quarterly holding from 2027; surrender by 30 SepVerifiedNet0; DEHSt certificates; CBAM Guide certificatesHighSets the per-tonne value of avoiding overstatement.
Default values carry a mark-up of +10% (2026), +20% (2027), +30% (2028); default-vs-actual gap can be ~80% (TR cement) to >130% (CN BF steel)VerifiedIR (EU) 2025/2621 via CBAM Guide default-values; Sustainability CloudHighThe recovery hook: verified actual data displaces marked-up defaults.
Actual emissions need an installation monitoring plan (IR (EU) 2025/2547) + third-party EU-accredited verifier (DR (EU) 2025/2551); on-site audit first year; 5% variance thresholdVerifiedCarbonChain; CBAM Guide; Zevero; Coolset; SustaliumHighDefines the licensed chokepoint we coordinate (and must not perform).
Supplier installation-level data collection is the single biggest operational bottleneck; fragmented, multilingual, multi-tier, manualVerifiedZevero; CarbonChain; CoolsetHighThe labor we automate & the reason a service beats a tool.
Incumbent budget: Big Four CBAM desks €15k–40k/quarter; software Dubrink ~€1,990/yr, CBAMBOO €9k–19k/yr; managed CBAM.in ~$499/quarter / $1,699/yrVerifiedFormist; Carbon Complete; Net Zero Compare; CBAM.inMed-HighProves an existing, redirectable budget & price anchors.
Per-declaration program fee + per-supplier data unit can reach 50%+ gross marginUnverifiedAuthor model; anchored to incumbent pricing aboveLowCentral economics — pilot kill-metric.
Specialist review compresses to minutes per supplier-installation at steady state once the dataset is reusedInferredInferred from rules-bound, template-driven nature of the workLowMargin lever — must be measured in pilot.
Mid-market importers will switch from DIY/Big-Four/software to a done-for-you AI-native deskUnverifiedNo direct buyer interviews this runLowDemand-quality assumption — validate before scaling.

Decisive selection rested only on Verified claims (binding obligation, penalties, certificate price, default mark-ups, verifier requirement, supplier-data bottleneck, incumbent budget). Every economic assumption is Inferred or Unverified and routed to the risk register and 90-day plan.

08 Why now

Verified shifts

The obligation just changed character: from a no-penalty quarterly report (Oct 2023–Dec 2025) to a financial regime with authorisation gating at the customs border, marked-up default values, and €100/t + €300–500/t penalties (Verified). The Omnibus simplification (Reg (EU) 2025/2083) deliberately concentrated the burden on larger importers, and the Commission has signalled a scope extension to downstream steel/aluminium-intensive goods and possibly new sectors in 2026 — meaning the in-scope buyer set is set to grow, not shrink (Verified, Council/Commission documents). More than 12,000 importers applied for authorisation in the first days of 2026 (Verified) — demand is here now, not hypothetical.

Inferred capability shifts

Frontier models now reliably draft multilingual supplier outreach, extract emissions data from heterogeneous supplier returns (PDFs, the EC communication template, mill certificates, spreadsheets), map goods to CN codes, production routes and precursors, and assemble structured declaration data — the exact extraction-and-coordination bottleneck that made CBAM compliance labor-bound. (Inferred from current document-AI capability; to be proven on real supplier files.)

Unverified hypotheses

That mid-market importers are dissatisfied enough with DIY/Big-Four/software handling to hand the whole job to a done-for-you desk, and that they will accept a default-avoidance savings share. (Unverified — buyer discovery is the first 90-day task.)

09 Customer & PMF

AttributeDetail
ICPEU importers of CBAM goods above the 50-tonne threshold with 0–1 in-house carbon-accounting specialists: steel stockholders/distributors & service centres, metal fabricators & OEMs importing semi-finished steel/aluminium, fertiliser & agri importers, cement/clinker buyers, and the indirect customs representatives (freight forwarders, customs agents) who hold declarant status for clients. Beachhead = iron & steel importers (dominant volume).
Economic buyerHead of Trade Compliance / Customs / Procurement, or CFO/Finance director who now owns the certificate line-item.
Champion / userCompliance or procurement manager currently chasing suppliers for emissions data in spreadsheets and email.
Urgent triggerAuthorisation just obtained; first quarterly 50%-holding obligation looms (from 2027); a supplier refuses/ignores data requests; finance sees the projected certificate bill at default values; scope extension brings new CN codes in-scope.
Alternatives today(1) DIY in spreadsheets + the EC template; (2) Big Four / boutique advisory (€15k–40k/quarter); (3) customer-operated software (CarbonChain, CBAMBOO, Dubrink, carbmee — €2k–31k/yr); (4) managed reporting services (CBAM.in et al.); (5) do nothing & eat default values + penalties.
Jobs-to-be-done“Get my suppliers' verified actual data so I stop overpaying on defaults”; “file a clean annual declaration on time”; “tell me how many certificates to hold each quarter”; “make the supplier chase someone else's problem”; “give me an audit-ready file if the competent authority asks.”
Willingness to payInferred from existing advisory (€15k–40k/quarter) and software-plus-service pricing; the default-avoidance savings share is self-funding when verified actual data is materially lower than default. Must be confirmed in pilot.

10 The outcome we sell

Deliverable

For each compliance year: (1) a complete annual CBAM declaration assembled and submitted in the CBAM Registry by 30 September (or prepared for the importer's/representative's submission), with embedded emissions per CN code and installation; (2) a run supplier-data campaign — installation identification, multilingual requests, follow-up, extraction, completeness scoring; (3) a documented default-vs-actual optimization driving verified actual data wherever it lowers the certificate bill, with the accredited-verifier engagement coordinated; (4) quarterly certificate-holding forecasts (the 50% rule from 2027) and a maintained, audit-defensible workpaper file.

Acceptance criteria

Declaration filed/ready before deadline (zero late filings); embedded-emissions figures reconciled to verified supplier data or correctly applied defaults; verified-actual coverage maximized for the highest-impact suppliers; certificate quantity reconciled to the declaration; an audit packet producible within 48 hours per declarant.

Customer promise

“Your CBAM declaration is filed on time, your suppliers' real data is in it wherever it saves you money, and you will not face the competent authority with a spreadsheet.”

Exclusions / refund-rework

We are not the accredited verifier and do not issue the statutory verification opinion; we do not provide legal/tax advice. If we cause a missed filing deadline, we cover the resulting penalty exposure within agreed limits and re-file at no charge. The default-avoidance share applies only to verified, realized certificate-cost reductions.

Measurable success metric: on-time filing rate (target 100%), verified-actual coverage rate (% of embedded emissions backed by verified supplier data), realized certificate-cost reduction vs. all-defaults baseline, and audit-packet turnaround.

11 Internal AI engine architecture

1 · Intake

Import data (customs entries / CN codes / net mass / country of origin), supplier & installation roster, prior CBAM reports, existing supplier emissions declarations, and the EC communication-template files — via secure upload or customs/ERP connector.

2 · Normalization

Parse heterogeneous supplier returns (PDF, Excel, the EC template, mill certificates); standardize installations, production routes, units; dedupe; reconcile declared tonnage to customs; version every data point with provenance.

3 · Retrieval / knowledge

Structured registry: CN-code-to-sector/route map, EU default values by country/sector, precursor relationships, MRR-aligned methodology (IR (EU) 2025/2547), verifier requirements (DR (EU) 2025/2551), deadlines and certificate-pricing rules — versioned and dated.

4 · AI workbench

Draft & send multilingual supplier data requests; extract emissions, activity data, and methodology tiers from returns; map goods to routes & precursors; draft default-vs-actual comparisons and the declaration line items; draft verifier-readiness checklists.

5 · Deterministic rules

Embedded-emissions math, default-value & mark-up application, 50% quarterly holding calculation, certificate-price application, threshold/deadline logic, and unit conversions run as code — never as model guesses.

6 · Human chokepoint

Carbon-accounting specialist approves methodology, system boundaries, route/precursor decisions, and default-vs-actual calls; coordinates (does not perform) the accredited verifier; signs off the declaration before filing. Operators clear extraction exception queues.

7 · QA

Pre-file completeness & reconciliation checks; verified-actual coverage review; cross-supplier variance and outlier detection; second-set-of-eyes on any installation above a tonnage/value threshold.

8 · Delivery

Assemble & submit the CBAM Registry declaration (or hand a registry-ready package to the declarant/representative); deliver the certificate-quantity reconciliation, savings summary, and audit packet.

9 · Learning loop

Verifier findings, competent-authority queries, supplier response rates, and rejected data feed back to improve outreach templates, extraction, route mapping, and the knowledge registry.

10 · Model portability

Model-agnostic extraction/drafting interface; swap or ensemble frontier models per task & language; deterministic layers and the verified supplier/installation dataset are model-independent assets.

12 AI-vs-human operations pipeline

AI

Ingest customs/import data & supplier roster; reconcile tonnage to CN codes.

AI

Identify installations; draft & send multilingual supplier data requests; chase.

AI

Extract emissions/activity data from returns; map goods to routes & precursors.

Rules

Apply EU defaults & mark-ups, embedded-emissions math, 50% holding & price logic.

Operator

Clear exception queue: unmatched installations, low-confidence extractions, missing fields.

Carbon specialist

Approve methodology, boundaries, default-vs-actual; coordinate accredited verifier; sign off.

Rules

Pre-file completeness, reconciliation & certificate-quantity checks; lock package.

AI

Assemble/submit registry declaration; capture proof; reconcile certificate holdings.

Customer

Receives filed declaration, certificate forecast, savings summary, audit packet.

Humans are concentrated at two chokepoints — operator exception-clearing and specialist methodology/verifier sign-off. Everything upstream, the supplier-chase, and the filing mechanics are AI- or rules-owned. The accredited verifier is an independent third party we coordinate, never replace.

13 Operations as product

  • SOPs per workflow: import intake, installation mapping, supplier-data campaign, extraction, default-vs-actual decision, declaration assembly, verifier coordination, certificate reconciliation — each a versioned runbook.
  • Structured intake checklist + automated completeness check before any declarant enters production.
  • Knowledge registry as the core asset: CN-code maps, EU default values, precursor relationships, methodology and verifier rules — versioned, dated, source-linked; refreshed each compliance cycle and at every implementing-act change.
  • Reusable verified supplier/installation dataset: once a mill's verified data is captured, it serves every importer buying from that installation (with consent) — a compounding, cross-customer asset.
  • Exception queues with confidence scoring: low-confidence extractions and unmatched installations route to operators; only true methodology/boundary calls route to the specialist.
  • Reviewer assignment logic by sector, language, declarant value, and emissions-impact tier.
  • Audit trail & version control on every data point, calculation, and filing — the audit packet is a byproduct, not extra work.
  • Gold-standard examples per sector/route; red-team checks for over-/under-statement and missed deadlines; root-cause & postmortem on every verifier finding or authority query, converted into a rule or registry update.

The product is the production system: experts improve the machine and handle the hardest methodology calls; the machine handles the supplier-chase volume and never forgets a deadline.

14 No-holes quality engine

Deadline integrity

Every declarant has a tracked annual-declaration deadline and quarterly 50%-holding checkpoints with escalation; a daily reconciliation proves “every declarant, every checkpoint, accounted for.” A looming miss is a P0 alert.

No over-statement (cost control)

Verified-actual coverage is maximized for the highest-impact installations; default values are applied only where actual data is genuinely unavailable, and flagged for next-cycle recovery.

No under-statement / integrity risk

Reconciliation to customs tonnage prevents omitting goods; large downward emissions swings vs. defaults require specialist sign-off, documented rationale, and verifier-backed evidence (defends against penalty exposure).

Hallucination control

Embedded-emissions, mark-up, holding, and price math are deterministic; model outputs are constrained to the knowledge registry; every emissions figure cites its source supplier document and methodology tier.

Verifier readiness

Data is assembled to the verifier's expected structure (monitoring plan, activity data, 5% variance tolerance); rejected items feed the learning loop. We never substitute our judgment for the accredited verifier's opinion.

Audit readiness

Every declaration carries linked workpapers and provenance; a competent-authority query is answered from the packet, not a scramble.

15 Pricing, pricing legality & unit economics

Model

Primary: an annual per-declaration managed program fee, tiered by number of CN-code sectors, installations, and supplier count, billed per compliance cycle and inclusive of quarterly certificate-holding forecasts. Per-unit: a per-supplier-installation data-collection fee for onboarding each new installation to verified-actual data. Optional: a default-avoidance savings share on verified, realized certificate-cost reductions. Why not hourly: hourly billing caps margin, punishes the customer for uncooperative suppliers, and rewards slowness — the opposite of an engine whose thesis is that each supplier-installation gets cheaper to onboard as the dataset and outreach templates compound.

Pricing legality

Fixed per-declaration and per-installation fees are unrestricted B2B service pricing. A savings share on certificate-cost avoidance is a commercial performance fee (not a regulated contingency on a legal recovery), but we will (a) define the all-defaults baseline transparently in the engagement letter, (b) tie payment to verifier-confirmed actual data to avoid disputes, and (c) avoid any structure that could be read as sharing in a tax/penalty outcome. We are a preparer/orchestrator, not the accredited verifier and not, by default, the authorised declarant of record; where we act as or alongside an indirect customs representative, that role's specific liabilities are handled under separate, explicit terms (§21). (Verified regulatory roles; pricing-structure legality validated with counsel in §27.)

Illustrative unit economics — onboarding one supplier-installation to verified-actual data (Unverified, model)

COGS driverEst. per installationNote
Model inference (outreach, extraction, mapping, drafting)€0.50–3.00Multilingual; falls as templates/models optimize.
Document processing / storage / hosting€0.30–1.00Per-installation amortized.
Operator exception-clearing & supplier follow-up€8–35Target falling as outreach automation improves; uncooperative suppliers higher.
Carbon-specialist review/sign-off€6–25Methodology/boundary & default-vs-actual calls only; routine flows through.
Verifier coordination (our admin; verifier fee billed to client)€3–12We schedule/package; the statutory verification fee is the client's pass-through cost.
QA + support + rework reserve€3–10Rework target <5%.
Total COGS / installation~€21–86Re-used across importers buying from the same installation; first-touch highest.

Targets: gross margin 55%+ at steady state; revenue/FTE €350k–650k as automation rises and the verified-installation dataset is reused. The accredited-verifier fee is a client pass-through, not our COGS. All figures are author estimates (Unverified) and are the central pilot kill-metrics.

16 Nonlinear scaling plan

LeverLaunch+90 days+1 year
Automation rate (installations onboarded without operator touch)20–35%45–60%70–82%
Operator minutes / supplier-installation30–6015–306–12
Specialist touch (% of installations)~100% (trust-building)30–45%12–22% (methodology/edge cases only)
Installations / operator / day4–1015–3550–100
Gross margin0–25%35–52%55–65%

Revenue decouples from headcount because the knowledge registry and the verified supplier/installation dataset are shared fixed assets: the same Turkish mill's verified data, once captured, serves every importer buying from it; the second compliance cycle for any declarant is far cheaper than the first (suppliers already onboarded, methodology already set). Margin expands as outreach/extraction automate and specialist touch falls to methodology 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, cheaper multilingual models make supplier outreach and emissions extraction faster and more accurate — directly expanding margin and throughput. The engine is architected model-agnostic, so we ride the curve. Score: 3.9.

What actually defends the business (because the methodology is public and software incumbents are funded): (1) the done-for-you accountable outcome — a customer keeps a vendor that files on time and lowers the certificate bill, not a tool that hands the supplier-chase back to them; (2) the reusable verified supplier/installation dataset, which compounds with every cycle and is expensive to rebuild; (3) the verifier-coordination relationships and supplier response infrastructure (templates, languages, persistence) that turn a fragmented chase into a system; (4) switching cost from a normalized multi-year emissions history living in our system; (5) proof-of-savings track record per sector/corridor.

Strongest commoditization threat: a funded software incumbent (CarbonChain, CBAMBOO, carbmee) or a Big Four desk adds an AI “managed-data” layer for the mid-market and bundles it with an existing customs/ERP footprint, compressing price before we reach dataset breadth. Anti-commoditization: even if general models make self-serve easier, the binding need for an accountable party, verifier coordination, and verified supplier data that the importer cannot extract alone remains — we win on outcome ownership and the proprietary verified-installation dataset, not model access.

18 Go-to-market

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

Why this GTM

Compliance & procurement leaders buy from credible specialists on referral and at trigger moments; they will not self-serve a penalty-bearing customs obligation across multilingual suppliers.

First 50 prospects

Authorised CBAM declarants (public/known set), steel service centres & stockholders, metal fabricators/OEMs, fertiliser importers, and freight forwarders / customs brokers acting as indirect customs representatives — concentrated in DE, IT, NL, PL, ES, FR.

Trigger events

Authorisation obtained; the projected default-value certificate bill hits finance; a key supplier ignores data requests; the 2027 quarterly 50%-holding rule approaches; CBAM scope extension adds new CN codes.

Outreach wedge

Free “CBAM Default Exposure Scan”: send us last year's import lines (CN codes, tonnage, origin); we estimate the all-defaults certificate cost vs. an achievable verified-actual scenario — a concrete euro number that funds the engagement.

Channel partners

Freight forwarders & customs brokers (declarant-of-record relationships), ERP/customs-software vendors, trade associations (steel/metals federations), chambers of commerce, and accredited verifiers (referral both ways — they verify, we prepare).

Conversion path & metrics

Scan → single-corridor managed pilot (one supplier set, one cycle) → full declaration program → multi-sector + quarterly forecasting. Track scan-to-pilot, pilot-to-program, on-time filing, verified-actual coverage, realized savings.

Credibility asset: a carbon-accounting specialist with EU MRR/CBAM methodology depth plus a published, verifiable savings case study per corridor. Answer-engine optimization: publish authoritative CBAM how-to content so the desk surfaces when importers research “CBAM declaration / default values / supplier data” in ChatGPT, Perplexity & Google. Expected sales cycle: weeks-to-a-quarter, accelerating near deadlines.

19 Pilot & early-demand-trap mitigation

Pilot cap

Max 5–8 declarants for the first cycle, deliberately concentrated in iron & steel from 2–3 supplier countries so the route/precursor logic and supplier dataset harden before breadth.

Pilot customer profile

One multi-supplier steel importer with uncooperative suppliers plus a couple of simpler single-corridor importers, to stress the engine across data-availability extremes.

Success criteria

100% on-time declaration; verified-actual coverage achieved on the highest-impact suppliers; documented certificate-cost reduction vs. all-defaults baseline; operator minutes/installation trending down; specialist touch falling below 100%.

Manual workarounds — tracked

Log every hand-fix (a supplier format the engine can't parse, a one-off route). Each becomes a template/registry/rule before scaling, never a permanent human patch.

What we refuse

No acting as the accredited verifier; no legal/tax advice; no bespoke carbon-strategy consulting that doesn't improve the engine; no sector/route we can't yet encode defensibly.

What kills the idea

If operator minutes/installation don't fall across cycles, or verified-actual data is too hard to obtain to fund savings, or specialist review can't drop below ~100% without quality loss — margins won't clear and we stop.

20 Competitive landscape

CategoryExamplesGap we exploit
Big Four / boutique CBAM advisoryBig Four CBAM desks, ex-DG-TAXUD boutiques€15k–40k/quarter, hourly-flavored, capacity-bound; mid-market is priced out or deprioritized.
Customer-operated CBAM softwareCarbonChain, CBAMBOO, Dubrink, carbmee, ClimEaseStill hands the supplier-chase & data entry back to the importer; we sell the outcome, not a tool to run.
Low-cost managed reportingCBAM.in (Dubai) and similar; one-time tools (CBAMCheck)Often default-values-only / report-generation; we own verified-actual data collection, verifier coordination & cost optimization.
Accredited verifiersEU-accredited verification bodiesNot competitors — partners; they verify, we prepare/orchestrate. Referral flows both ways.
In-house compliance teamsSpreadsheets + EC template + emailDrowning in multilingual supplier chasing; we remove the burden & the over-payment.
Do nothingDefault values + mark-ups, €100/t & €300–500/t penalties, goods stopped at the border.

21 Regulation, compliance & licensing boundary

What AI/operators may do

Run supplier outreach, extract & structure emissions data, map CN codes/routes/precursors, apply EU defaults & methodology, compute embedded emissions, draft the declaration, forecast certificate holdings, and prepare workpapers.

What an independent accredited verifier must do

Perform the statutory verification of actual emissions (monitoring-plan check, on-site audit in year one, 5% variance tolerance) under DR (EU) 2025/2551. We coordinate and package for them; we never issue the verification opinion.

Declarant-of-record boundary V

The authorised CBAM declarant (the importer, or an indirect customs representative acting for it) is the legally responsible filer. By default we prepare and the client/representative submits; if we partner with or act as an indirect customs representative, that role's liabilities are governed by separate explicit terms.

Prohibited claims

No guarantee of a specific emissions figure or certificate count; no understating emissions; we never advise omitting goods. We do not hold ourselves out as an accredited verifier or as a provider of legal/tax advice.

Privacy & data

Supplier emissions data may be commercially sensitive; handled under NDAs/DPAs, encryption, access controls, audit logs, and consent before any cross-importer reuse of a verified installation's dataset (GDPR where personal data is involved).

Per-scope expansion gate

Before adding a sector/corridor, confirm the current implementing acts, default-value tables, and verifier availability, and encode them into the engagement template and registry.

22 Founding team & expert map

RoleWhy neededFT / fractionalFirst hire
CBAM / carbon-accounting specialist (EU MRR methodology)Methodology sign-off, verifier coordination, credibility & customer interfaceFull-timeFounder / first hire
Ops leadOwns SOPs, supplier-data campaign, exception queues, filing calendarFull-timeMonth 1
AI/automation engineerOutreach, extraction, knowledge registry, declaration assemblyFull-timeFounder / first hire
Customs / trade-compliance advisorDeclarant-of-record & indirect-representative rules, customs dataFractional → FTPre-pilot
Sales / channel leadOutbound + forwarder/broker & verifier partnershipsFractional → FTPost-pilot
QA ownerPre-file checks, verified-actual coverage, audit-packet integrityFractional (ops doubles early)As volume grows

23 Exhaustive risk register

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

If uncooperative, multilingual suppliers keep operator + specialist minutes high, gross margin never clears 50%. Mitigation: measure minutes/installation obsessively in pilot; automate outreach & extraction relentlessly; reuse verified-installation data across importers; price first-cycle onboarding separately from steady-state filing. Owner: Ops + AI lead. Leading indicator: minutes/installation trend cycle-over-cycle.

2 · Funded software incumbent adds a managed-data layer and compresses price
Likelihood: Med-HighImpact: HighVerified threat

CarbonChain/CBAMBOO/carbmee or a Big Four desk could bundle AI managed data into the mid-market. Mitigation: compete on accountable outcome, verifier coordination, verified-installation dataset, and mid-market focus; prioritize dataset breadth and switching-cost data; partner rather than fight where possible. Owner: Founder. Leading indicator: incumbent mid-market pricing & win/loss.

3 · Suppliers refuse to provide verified actual data
Likelihood: HighImpact: HighVerified pain

Non-EU suppliers lack systems/incentive; verifier capacity is tight. Mitigation: persistent multilingual outreach + supplier education kits; prioritize highest-impact installations; fall back to correctly-applied defaults with a documented next-cycle recovery plan; build supplier-side onboarding so verified data becomes a reusable asset. Owner: Ops. Leading indicator: supplier response & verified-coverage rate.

4 · Default-vs-actual savings smaller or more contested than assumed
Likelihood: MedImpact: HighInferred

If a supplier's actual intensity is near the default, the recovery hook weakens. Mitigation: don't over-index on the savings share; the compliance value (on-time filing, deadline integrity, audit defense, certificate forecasting) stands alone. Owner: Carbon specialist. Leading indicator: measured savings per pilot declarant.

5 · Missed filing / holding deadline caused by us → penalty + reputational hit
Likelihood: Low-MedImpact: HighVerified penalty

€100/t under-surrender; €300–500/t unauthorised. Mitigation: deadline integrity as a P0 system with daily reconciliation, redundant alerts, and the quarterly 50%-holding tracker; we cover penalties we cause within agreed limits; E&O insurance. Owner: Ops. Leading indicator: deadline-reconciliation exceptions.

6 · Verification fails or actual data rejected (5% variance / methodology error)
Likelihood: MedImpact: HighVerified requirement

If verified-actual data is rejected, the declarant falls back to punitive defaults at the worst moment. Mitigation: assemble strictly to verifier expectations; pre-check against the 5% variance tolerance; maintain methodology QA; keep a defensible default fallback ready. Owner: Carbon specialist. Leading indicator: verifier finding/rejection rate.

7 · Regulatory drift (implementing acts, default tables, scope extension change yearly)
Likelihood: HighImpact: MedVerified churn

CBAM is actively being amended (Omnibus 2025; downstream extension proposal 2026; new implementing acts). Mitigation: registry versioned with a pre-cycle refresh sprint and change-detection on EUR-Lex/EC sources; scope extension is upside (more in-scope buyers). Owner: AI + Carbon specialist. Leading indicator: registry change-log cadence.

8 · Declarant-of-record / indirect-representative liability ambiguity
Likelihood: MedImpact: Med-HighVerified (roles)

Mitigation: default to preparer/orchestrator with the client/representative as filer; only assume indirect-customs-representative status under separate explicit terms and insurance; counsel review per jurisdiction. Owner: Customs advisor + counsel. Leading indicator: legal sign-off per engagement type.

9 · Supplier data quality / heterogeneity blocks automation
Likelihood: Med-HighImpact: MedInferred

Returns arrive as inconsistent PDFs/spreadsheets in many languages. Mitigation: robust normalization layer; standardized request templates that constrain supplier output; structured intake checklist; price first-cycle cleanup. Owner: AI + Ops. Leading indicator: extraction exception rate.

10 · Buyers won't fully outsource without a crisis
Likelihood: MedImpact: HighUnverified

Mitigation: lead with the free Default Exposure Scan (a euro number); target trigger events (authorisation, looming certificate bill, supplier refusal); white-label through forwarders/brokers. Owner: Sales. Leading indicator: scan-to-pilot conversion.

11 · Seasonality / deadline concentration (annual + quarterly checkpoints)
Likelihood: HighImpact: MedInferred

Work clusters around 30 Sep and quarterly 50%-holding dates. Mitigation: automation removes the labor peak; stagger supplier onboarding through the year (data collection is continuous, not just at deadline); smooth revenue with quarterly forecasting retainers. Owner: Ops. Leading indicator: peak-period utilization.

12 · Political / scope retrenchment (CBAM weakened or trade-deal carve-outs)
Likelihood: Low-MedImpact: Med-HighInferred

Trade negotiations or future simplification could narrow the regime; political salience is high. Mitigation: treat as multi-year, not existential; diversify sectors/corridors; the verified-supplier dataset and customs-data muscle are reusable for adjacent regimes (UK CBAM from 2027, product-carbon-footprint rules). Owner: Founder. Leading indicator: regulatory/political signals & in-scope CN-code count.

24 Tech stack & build plan

Stack

Python services; frontier LLM(s) behind a model-agnostic extraction/drafting interface (multilingual); document parsing (PDF/Excel/EC template); a structured knowledge registry (Postgres, versioned) for CN maps, default values, precursor relationships, methodology & verifier rules; deterministic rules engine for embedded-emissions/mark-up/holding/price math; secure object storage with audit logging; supplier portal + email-based outreach automation; customs/ERP connectors + secure upload; CBAM Registry submission integration / registry-ready export; internal ops console with exception queues & reviewer assignment.

Build sequence

(1) Import-data intake + tonnage/CN reconciliation. (2) Knowledge registry for iron & steel (routes, precursors, default tables). (3) Supplier-data campaign engine (templates, outreach, follow-up, extraction). (4) Embedded-emissions calc + default-vs-actual comparison (deterministic). (5) Declaration assembly + pre-file QA + certificate-holding forecaster. (6) Verifier-coordination & packaging workflow. (7) Learning loop + multi-sector registry expansion (aluminium, fertilisers, cement).

No “use agents” hand-waving: discrete services with explicit inputs/outputs, deterministic math in code, model calls only for outreach/extraction/drafting under specialist review; the statutory verification stays with an external accredited body.

25 Metrics & KPIs

Throughput

Installations/operator/day; declarants/cycle.

Cycle time

Import list → declaration ready; supplier request → verified data.

On-time filing rate

Target 100%; zero self-caused late/under-surrender.

Verified-actual coverage

% of embedded emissions backed by verified data.

Realized certificate-cost reduction

€ saved vs. all-defaults baseline.

Automation rate

% installations with no operator touch.

Specialist touch rate

% installations needing methodology sign-off.

Rework / quality-failure rate

Target <5%; verifier rejections trending down.

COGS/installation & gross margin

Margin target 55%+ at steady state.

Revenue/FTE

Target €350k–650k as automation rises.

Supplier response rate

Health of the data campaign.

Scan→pilot→program conversion

Funnel health.

26 What could kill this

  • Margins never clear 50% because uncooperative multilingual suppliers keep review minutes high — the fastest, most likely killer.
  • Suppliers structurally refuse verified-actual data, leaving everyone on defaults and gutting the savings hook.
  • A funded software incumbent bundles AI managed data into the mid-market at a price we can't match before we reach dataset breadth.
  • Buyers won't fully outsource off DIY/software absent a crisis, making CAC and sales cycles unworkable.
  • A self-caused missed deadline / under-surrender early on craters trust in a reference-driven, penalty-bearing market.
  • Political retrenchment (trade-deal carve-outs or further simplification) materially narrows the regime before we diversify.

27 90-day validation & launch plan

WeeksFocusActions & evidence gaps to close
1–2Buyer discovery15–20 interviews with trade-compliance/procurement/finance leaders at authorised declarants & steel importers; confirm pain, outsourcing appetite, pricing tolerance (closes the biggest Unverified gaps).
2–4Legal & role boundaryCounsel review of declarant-of-record / indirect-customs-representative liability and savings-share legality in beachhead states; finalize engagement-letter templates & verifier-partner referral terms.
3–6Engine MVPBuild import intake + iron & steel knowledge registry + supplier-data campaign engine + extraction + deterministic emissions/default-vs-actual calc.
5–8Free-scan wedgeRun Default Exposure Scans on 5–10 prospects' real import lines; quantify all-defaults vs. achievable-actual cost (closes the recovery-hook gap).
6–10Pilot recruitmentSign 5–8 pilot declarants (cap enforced); instrument minutes/installation, automation rate, specialist touch, supplier response from day one.
8–12Pricing testTest per-declaration + per-installation + savings-share vs. fixed-only on pilots; validate margin trajectory.
10–13Compliance & QA hardeningStand up deadline-integrity + quarterly-holding tracker, verifier-readiness packaging, second-eyes review; document SOPs; secure an accredited-verifier partner.
OngoingKill criteriaStop if minutes/installation don't fall across cycles, verified-actual data can't be obtained to fund savings, specialist touch can't drop below ~100% without quality loss, or no pilot converts to program.

28 Sources

The definitive-regime start (1 Jan 2026), the annual-declaration / 30-Sep-2027 deadline, the authorised-declarant requirement, the 50-tonne de minimis (Reg (EU) 2025/2083) and its ~90%/99% effect, the >12,000 applications / 4,100+ declarants and iron-&-steel dominance, the €89bn / ~105 Mt 2024 footprint, the €100/t and €300–500/t penalties (Art. 26 as amended), the €75.36/t Q1-2026 certificate price, the default-value mark-ups (+10/+20/+30%, IR (EU) 2025/2621), the verified-actual + accredited-verifier requirement (IR (EU) 2025/2547; DR (EU) 2025/2551; 5% variance; first-year on-site audit), the supplier-data bottleneck, and the incumbent pricing are Verified via the sources above (retrieved 2026-06-30). Per-declaration/per-installation fees, COGS, gross margin, revenue/FTE, automation rates, review-minute compression, default-vs-actual savings magnitude, and outsourcing willingness 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.