The Viictor Approach

How AI should work in customs & trade compliance.

Viictor is an AI colleague grounded in the customs framework itself — not a general AI that has read about customs. It reasons over the standards you already work with — WCO, UN/CEFACT, EUCDM, the trade agreements you file under every day — not over the whole internet.

Most systems automate a procedure. This one holds the relationships: a classification isn’t a step in a workflow, it’s a link between a product, a rule, a ruling, and the person who signed it off.

Two decades of customs systems. ISO 9001. Cyber Essentials.

Grounded in the framework you already work with.

The difference between a plausible answer and a correct one, in customs, is the framework the answer is built on. Viictor doesn’t invent that framework — it’s grounded in the standards the profession already relies on. What Viictor does invent is the reasoning that turns those standards into answers you can defend.

  • WCO Data Model — the international structural language of customs

  • UN/CEFACT standards — cross-border trade data and processes

  • EUCDM — the EU Customs Data Model, mapped to our declaration taxonomy across the declaration types in scope

  • Trade agreements — Protocol 4, the PEM Convention, EU FTAs, and the specific rules of origin they carry, kept current

  • Regulatory codes — TARIC, HS, EORI — the reference registries that anchor every declaration

  • Regulations in force — the UCC (Reg. 952/2013 + Reg. 2015/2447) and their derived acts

  • Curated customs knowledge — a global base assembled and maintained by customs practitioners — not by web crawl — that your own documents then specialise, per organisation

  • Workflows — the procedures a customs professional actually follows, encoded as executable steps

Traceable by design

Every answer Viictor produces is traceable to the source that grounds it. When it tells you the duty rate under EU-Morocco preference, it cites Protocol 4 (OJ L 70/2000) and the PEM Convention (OJ L 54/2013). When it flags a classification, it cites the HS heading and any Explanatory Notes it consulted. When it recommends applying for a Binding Origin Information, it names the reason.

That grounding is what separates an AI that reasons from an AI that talks. A general-purpose model can tell you what a rule of origin sounds like. Viictor tells you what the rule of origin is, for your goods, under your trade agreement — and shows you the sources, so your decision is defensible.

Not just AI that talks. AI that reasons.

“Couldn’t I just ask ChatGPT or Claude?” Even for a quick customs question, Viictor Free already answers it better and costs nothing because it is grounded in the framework rather than in whatever the web happens to say. And for a decision you have to defend to an auditor, the gap widens: a general assistant guesses in the gaps, can’t see your live operation, and can’t prove why it’s right. Viictor runs the same class of AI, but wraps it in a formal model of your operation and the rules — so answers are checked, traced and defensible, not just fluent.

How a general AI assistant compares with Viictor across nine compliance situations.
Situation A general assistant (e.g. ChatGPT, Claude) Viictor
Your operation Briefed from scratch every chat — you paste the context, and it forgets. Already models your products, parties, trade-lanes and authorisations — held live in your own ecosystem.
Your own knowledge The open web, plus whatever you paste in each time. Global customs knowledge from our practitioners, specialised by your own documents — private to each organisation you work for.
Right vs. sounding right Optimised to sound convincing; fluently wrong in the gaps. Checked against a validated rule set — correct, or it tells you it can’t yet be sure.
When it isn't sure Fills the gap with a confident guess. Stops, says what’s missing, and won’t act until it checks out.
Proof Prose you have to verify yourself. A trail to the exact rule and the exact fact in your data — auditor-ready.
Currency & scope A generalist with a training cut-off, answering anything. Customs & trade only, with the rule base kept current (WCO, UN/CEFACT).
Where you stand Answers the question you asked. It has no view of your compliance position. Assesses your own procedures against every checkpoint, names each gap, and re-assesses as your knowledge grows.
Work that outlives a chat Can group chats in a folder. It can’t hold an objective, a plan, or who owns what. A compliance project holds the objective, strategy, stakeholders and course of action — and every chat inside it starts briefed.
From answer to action Gives you words; you go do it in your systems. Carries the decision into your stack — staged, with your confirmation.

Three roles, one colleague.

Customs work isn’t one job. It’s expert guidance when a question comes up, careful configuration when a new regime needs to be set up, and steady operation once the flows are running. Viictor works in all three modes.

Viictor as a Consultant · VaaC

Expert customs guidance, on your data.

The advisory role. You ask a question — a supplier declaration that doesn’t add up, a new origin rule that dropped mid-shipment, a classification that’s costing you the wrong duty rate — and Viictor works the problem the way a senior customs professional would.

It reads the rule. It looks at the evidence. It calls out what it’s not sure about, and what would make it sure. It cites the sources. It tells you the confidence level.

A general-purpose model sounds right; Viictor is defensible — with a reasoning trail.

See the Consultant →

Viictor as an Assistant · VaaA

Configure your operations, in plain conversation.

The setup role. You tell Viictor what you want to monitor — a trade lane, a product family, a regulatory regime — and it proposes a compliance workflow, in plain language, that you can review, adjust and approve before anything goes live.

Every change is staged. Every commitment is gated on your confirmation. Nothing is mutated in your systems without a human reading it, understanding it, and saying yes.

AI proposes; you dispose; the record shows both.

See the Assistant →

Viictor as an Operator · VaaO

Watch live trade, catch what breaks, drive the fix.

The run-time role. Once a workflow is configured, Viictor runs it. It monitors your live shipments, watches for the conditions that mean something is going wrong — a missing document, a preferential rate at risk, an origin certificate that hasn’t arrived — and drives the fix through your existing systems.

When a decision exceeds Viictor’s authority, it escalates. When a routine action is safe, it proceeds. Every event is logged. Every audit trail closes.

Real-time monitoring and automation — never a black-box agent on your production trade.

See the Operator →

One colleague, three modes. You start where you need to start — and most teams need all three, because the advice, the setup and the operation are the same job seen from three angles, and Viictor keeps them coherent.

Six principles behind Viictor.

Six commitments about how AI should behave in customs and trade compliance. They shape every design decision we make.

  1. 1

    Built for the customs professional.

    Viictor is a colleague, not a construction kit. It arrives with the framework already loaded, the standards already current, the workflows already shaped by people who file declarations for a living. You don’t need to be a data scientist to put it to work; you need to be a customs professional who wants to be supported.

    Turnkey capability, from the first conversation.

  2. 2

    One governed view of every declaration, party and trade lane.

    Customs data is spread across ERPs, TMSs, broker portals, spreadsheets and emails. Viictor doesn’t ask you to consolidate it before you can use AI on top of it. It orchestrates across what you already run, keeps a single governed view of every declaration, every party and every trade lane, and produces intelligence from the connections between them — not just from the individual data points.

    Cross-domain visibility. No silos. No re-platforming.

  3. 3

    Configuration, not code, when the regulation changes.

    Customs regulation moves. New free trade agreements enter into force. New restrictions land. Trade lanes reroute. Viictor is built so that adapting to a change is a configuration act — new rules, new workflows, new evidence requirements — not a software project.

    The workflows are declarative. The knowledge base is versioned. The changes take effect where they need to, with the audit trail that shows what changed, when, and by whom.

    Ready for the unknown, without a six-month integration plan.

  4. 4

    AI advises. The customs professional decides.

    This is the principle we hold most firmly. Every determination Viictor produces — a classification, an origin ruling, a valuation, a coverage assessment, a release recommendation — is advisory to the professional who is legally accountable for the decision. Viictor tells you what it reasoned. It tells you what it cites. It tells you what it isn’t sure about. It never tells you what to do without showing you the working.

    That’s not a caveat added at the end — it’s designed into every role: the Consultant advises; the Assistant proposes and stages; the Operator acts only within a scope you approved, and escalates the moment it’s outside that scope.

    There is a person whose signature matters. Viictor makes it better informed — it does not replace it.

  5. 5

    Restrictive by default. Auditable exceptions.

    Access to customs data is not a general-purpose problem. It is a compliance problem with legal, jurisdictional and commercial constraints. Viictor is built restrictive by default: minimum data, minimum scope, minimum surface. Where an exception is warranted — a broader authority, a cross-jurisdictional view, an emergency override — it happens as an explicit, audited event, not as a permission drift.

    Every exception an explicit, audited event. Every derived work carries its provenance.

  6. 6

    Compliance built into the platform, not added later as a reminder.

    Compliance is not a notification you send after the fact. It is a property of the way the platform is built. Viictor’s data model is compliance-shaped: every consignment carries its lineage; every derived value carries its confidence; every rule is versioned; every workflow is declarative and auditable end to end.

    You don’t remind Viictor to be compliant. Being compliant is what Viictor is.

See it reason on your own operation.

Bring a question you’d want an auditor to accept — a classification, an origin call, a supplier declaration — and see the difference between an answer that sounds right and one you can defend.