A working prototype: from supplier files to an HS / TN VED candidate a person would have to sign.
We normalize the inbound pack, prepare a reviewed name and goods description, mark gaps and conflicts, and propose a code candidate with a rationale you can follow. This is a demonstration of the method — not a deployment for brokers, and not a bot that “declares itself”.
- A supplier pack almost never arrives in the form an EAEU importer needs. That is the prototype’s problem, not a description of someone else’s deployment.
- An expensive specialist spends days assembling someone else’s pack. The prototype prepares the record. The signature stays with a person.
- A code and a goods description on a declaration have to survive review — that is the specialist’s job, not the model’s.
- Certification rules are not in the spreadsheet.
not enough evidence for a full code
Normalize · Enrich · Validate · Classify · Explain · Escalate
In the prototype, EAEU TN VED classification is a candidate with a rationale. The decision is not automatic.
Hand-off to a person is in the architecture, not a footnote.
- Is this autonomous filing?
- No. Scanovich does not file and does not replace the customs broker.
- When invoice and packing list disagree?
- We treat the mismatch as a task of its own — we do not simply take the newer file.
- When do you see data?
- A request first. Files move only after a reply, on an agreed channel off this site.
- Who signs the code?
- The specialist of whoever files. Scanovich has no such customers today.
- Do you replace the broker?
- No.
If the query was “AI for customs brokers” — we do not replace the broker and we do not file. This page is a prototype of the trade work, not a deployment. We do not claim ranking in search models.
An internal set used to test the method: not a customer engagement and not accuracy on live declarations.
Evaluation.
Leave an email or Telegram. This is a conversation about a technical evaluation, not a desk and not customs filing. No files on this step. We do not claim completed deployments.