Episode 11: The Global Customs Canon: Common Trade Data Before AI in Customs
Former Head of Global Customs Programs - DHL
Greg Pilkington is a global customs and trade transformation leader specializing in customs modernization, AI architecture, and cross-border compliance. As Head of Global Customs Programs at DHL Express, he leads customs technology initiatives across more than 220 countries and territories, integrating data quality, automated clearance, and compliance risk management. He is also the founder of the Global Customs & Trade Forum and creator of the Global Customs Canon, a comprehensive doctrine for modernizing customs systems through data architecture, emerging technologies, and regulatory alignment. His experience includes senior strategy and management roles at the World Customs Organization.
Greg Pilkington, who led global customs programs at DHL Express and founded the Global Customs and Trade Forum, explains his Global Customs Canon. His argument is that AI in customs will stall until countries and vendors share one data language. He describes the trade line assertion, the levels of data fidelity from line items to verifiable credentials, a bilateral digital twin that could halve declaration work, and why visibility on a blockchain is not what every trader wants.
5 Key Findings
-
Structured data that disagrees is still chaos
AI can structure messy trade documents, but each vendor and jurisdiction produces a different structure. Without a shared ontology and canonical messages, the two data sets do not map and people are back to reconciling them by hand.
-
The trade line assertion links line, piece and document
Greg's smallest unit of trade data ties each commercial invoice line to the package it sits in and the certificates or licenses it needs. Today that link exists only in a broker's head.
-
Piece-level data stops the missing-box dispute
A common fraud pattern: five pieces declared, four arrive, and the missing one held the valuable goods. GS1 barcodes, EPCIS events and IoT tags make piece-level reconciliation practical and are already in use.
-
A bilateral digital twin removes repeat work
If the exporting country builds a verified data record that the importing country accepts under a plurilateral agreement, data entry at destination and in transit countries largely disappears. Audits and spot checks would still continue.
-
Share only what each party needs
Greg compares customs to card payments: the restaurant never sees your balance. Distributors avoid systems that expose their suppliers, so federated data, with each authority seeing only what it requires, matters more than full visibility.
Advice to the Professional
From Greg Pilkington
Customs has reached the C-suite, so stop treating it as a back-office task. Regulatory change will keep speeding up as technology allows, and brokers are already using AI classification heavily. Greg’s framework means professionals need to understand data models and verifiable credentials alongside classification, valuation and origin.
Start where you are with level-one data: clean ERP line items with the right HS code, unit of measure and duty rate, matched to the right piece. Move from document-first habits to data-first thinking, and join communities where these standards are being discussed.
Conclusion
Greg sees a common data language as the precondition for AI in customs. If countries adopt shared definitions and trusted records, trade could move faster and become harder to divert or smuggle.