The eFTI framework assumes structured data. A large share of real shipment documentation is not structured — invoices, packing lists, ADR reports and manifests arrive as documents, and someone has to read them. Compliance-as-a-Service attacks that gap directly: AI-assisted extraction turns unstructured shipment documents into structured, semantically validated data that competent authorities can use inside the eFTI framework.
The design point that distinguishes this pilot is what it refuses to automate. Extraction and validation are automated; the decision is not. Competent authority officers remain the decision-makers, and the pilot’s own success criteria include explainability, traceability and auditability of every output — so an officer can see why a value was extracted and where it came from. Governance and lineage tooling sits alongside the extraction engine for exactly this reason.
Operationally the pilot ingests shipment documents at a port or hub, extracts goods descriptions, quantities, origin and actors, validates completeness and consistency, and delivers ADR reporting for authority review. It also tests whether this holds up at realistic document volumes rather than on curated samples. It is a new pilot, introduced after Deliverable D2.1 and run in collaboration with Collibra.
