Another follow up on Ontology from the DMs worth sharing:

Q: It's rare to see this problem framed at the ontological level rather than as just another document, data, or software issue. The real shift isn't from paper to digital documents, it's from documents as primary objects to execution facts as primary objects.

A: That framing is exactly right. And it's the Canon's founding premise.

The document is not the primary object in trade.

The execution fact is: the assertion that a specific invoice line, representing a specific physical piece, is governed by a specific regulatory credential.

That relationship is what customs authorities need to reason over. The document is how that relationship has historically been packaged for transmission. It is not the thing itself.

Legal documents are derived outputs.

This isn't a provocative reframe.

It's an architectural one, and the implications for anyone building AI in customs are significant.

If your data model is built around documents, your AI is reasoning about the packaging, not the substance.

The Canon's Trade Line Assertion sits at L1 of the knowledge graph precisely because of this. OWL 2, JSON-LD, SHACL throughout. The graph asserts the relationship between invoice line, physical piece, and regulatory credential. Legal documents are generated from that graph on demand. The direction of dependency is reversed.

What's notable is that serious foundational work elsewhere in the field is arriving at the same architecture independently, pre-domain languages and formal algebras for trade execution facts, the same formalisation stack. The convergence is not coincidental. It reflects the underlying structure of the problem.

For those building or buying customs AI, the question is not which model you're running. It's what your data model is asserting, and whether it's asserting the right things about the right objects.

The DMs are open.