The BACUDA Project has contributed an article, “Data Governance as the Foundation for AI in Customs,” to the Herald published by the WCO RILO ECE, Issue 2/2026. Drawing on recurring challenges observed in BACUDA’s capacity building activities, the article argues that AI and data analytics can deliver reliable results when a robust data-governance foundation is already in place, and sets out six essential pillars for building such a foundation. The full text is available below.

Data Governance as the Foundation for AI in Customs

By Hyunsang Jo, the WCO Bacuda Project Manager

Interest in artificial intelligence (AI) and Data Analytics is growing rapidly across the Customs community. From risk management to revenue collection and fraud detection, many Members have already begun deploying AI and machine learning in their operations, and requests for related capacity building activities continue to increase. The 2024 WCO Smart Customs Survey highlighted the same trend, with Members identifying Big data and Data Analytics, AI/ML and blockchain as the technologies they were most likely to adopt.

However, advanced data analytics and AI can deliver meaningful results only when robust data governance is already in place.

This is a recurring challenge in the BACUDA Project’s capacity building activities. Many Members seek to move directly to the adoption of analytical models without first establishing the necessary data-governance foundations. Introducing AI without effective governance is much like constructing a building without a blueprint. When data is inaccurate, incomplete or inconsistent across systems, even the most sophisticated model will struggle to produce reliable results. This is captured by the familiar phrase “garbage in, garbage out (GIGO)”. Budget constraints, changes in senior management and the absence of a governance framework make progress even more difficult.

The same concern has been reflected in recent discussions at the WCO. When reviewing the outcomes of the 2026 WCO Technology Conference and the 94th Session of the Policy Commission, it was noted that AI had moved from aspiration to practical application. Nevertheless, it remained a tool rather than a solution, with its effectiveness entirely dependent on the quality and governance of the underlying data.

Earlier, at the 251st/252nd Sessions of the Permanent Technical Committee (PTC), Members emphasized that the successful adoption of disruptive technologies depended not only on the technologies themselves but also on governance, interoperability, cybersecurity, data quality, institutional readiness, human oversight, skills and realistic implementation approaches.

What, then, needs to be in place?

Data governance can be viewed as the blueprint, while data management is the contractor that builds according to it. Construction may begin without a blueprint, but the process will be inefficient, and the finished structure is likely to contain serious defects and compliance risks.

To avoid this such outcomes,  six pillars are essential: (i) strategy and policy, which set the overall direction; (ii) organization and roles, which establish clear responsibilities and accountability; (iii) standards and metadata, which provide a common language and shared understanding of data; (iv) data quality, which  ensures that data is accurate, complete, consistent and trustworthy; (v) security, privacy and ethics, which protect both the data and the trust associated with its use; and (vi) monitoring and compliance, which ensure that the governance framework remains effective over time.

If any one of these pillars is missing, the structure will be weakened.

WCO BACUDA Project places strong emphasis on data governance throughout its capacity building activities, including workshops and scholarship programmes. It has also launched a new e-learning module, Data Governance for Customs, now available to all WCO Members on the WCO CLiKC! platform.

As the use of AI and data analytics continues to expand, Members are encouraged to devote equal attention to establishing a strong and sustainable data-governance foundation. Without it, AI may generate outputs, but it cannot consistently generate reliable, responsible and actionable results.