At the Global Trade Leadership Summit in September, AMD gave delegates a practical look at how artificial intelligence is being used within its global trade compliance function.
The session, delivered by Declan Ivory, Director of Global Trade Compliance for Europe, Middle East and India, focused on a simple question: where can AI genuinely help compliance teams manage risk?
The answer, according to AMD, is particularly relevant in two areas: who we trade with and what we declare.
One of the strongest examples was customer and sanctions screening. Trade compliance teams can be dealing with information scattered across corporate records, sanctions lists, customs data, ownership information, trade routes, adverse media and other sources. AMD is using AI to help bring these different signals together and identify potential risks.
The aim is not for AI to make the decision. Instead, it can help a compliance professional get to the relevant evidence more quickly and build a clearer picture of the customer.
The same principle applies to export-control due diligence. AI can assist with researching end users and end uses and identifying information that may indicate diversion or prohibited activity.
AMD highlighted the use of AI to compare broker declarations against approved classification, origin and value data. Where differences are identified, they can be reviewed and corrected, helping improve the underlying data as well as the immediate declaration.
Perhaps the most useful part of the presentation was AMD’s discussion of where AI can go wrong. AI can produce incorrect information with complete confidence. It can miss important jurisdictional differences and apply a generalised answer where a specific legal framework is required.
AMD shared that some of its KYC prompts run to more than a page, specifying where the system should look, what it should include or exclude, and how the findings should be presented. The prompts can also go through several iterations before the team is satisfied with the result. For anyone working in trade compliance, this is an important point. Getting useful results from AI requires considerably more than asking a quick question and accepting the answer.
AMD’s approach can be summed up in three words: Trust and Verify. The company highlighted three important guardrails: approved tools, named human sign-off and retained records. AI output remains a draft until it has been reviewed and validated by the appropriate person. AI can search faster. It can connect information. It can identify patterns and help organise evidence. But the responsibility for the compliance decision still sits with the qualified professional.
For trade teams considering how to introduce AI into their own work, AMD’s presentation offered a useful starting point: look first at the areas where the evidence burden is highest, use AI to support the process, and build verification into the workflow from the beginning.
The technology is moving quickly. For global trade professionals, understanding how to use it responsibly is becoming part of the job.
