Concept Piece AI in Trade Compliance

Agentic AI in Trade Compliance: What the Word Actually Means

Last reviewed:

Agentic means a system that acts, not one that answers. Gartner found just 15% of IT leaders pursuing fully autonomous agents — governance, not capability.

Chapter 1

The Word Is Doing Two Jobs

"Agentic" has become the word vendors reach for when "AI" stops impressing, and it has drifted far enough from its technical meaning that buyers should insist on a definition before a demonstration. Stripped of marketing, an agentic system is one that pursues a goal by taking a sequence of actions in the world, choosing those actions itself, rather than producing an answer and stopping. The difference from a conversational assistant is not intelligence. It is agency — the system does things, and the things it does have consequences that persist after the session ends.

In most enterprise contexts that distinction is a matter of engineering risk. In cross-border trade it is also a matter of law, because the actions available to an agent in this domain — transmitting a declaration, lodging a notification, submitting a claim document — are acts that regulation attributes to a specific accountable party. An agent that merely drafts has produced a document; an agent that transmits has performed a regulated act in someone's name.

The market has already discovered this, quietly. Enterprises are adopting agents broadly and autonomy narrowly, and the gap between those two numbers is the most informative datum available about where this technology actually stands.

1.1

What the Adoption Data Shows

Enterprise appetite for agents is high; enterprise appetite for unsupervised agents is not. Gartner's survey of 360 IT application leaders at organisations of at least 250 employees across North America, Europe and Asia/Pacific, conducted in May and June 2025, found that 75% were piloting, deploying, or had deployed some form of AI agent — while only 15% were considering, piloting or deploying fully autonomous agents, which Gartner defines precisely as "goal driven AI tools that do not require human oversight."

Get the details
1.2

The Legal Ceiling on Autonomy

Autonomy in trade compliance is capped by law before it is capped by capability, and the cap is explicit in the European framework. Article 14(1) of the EU AI Act requires that high-risk AI systems be designed and developed so that they can be "effectively overseen by natural persons" during the period in which they are in use. This is a design obligation, not a deployment preference: a system architecture that makes meaningful human oversight impractical is non-conforming irrespective of how well it performs.

See more about
Get the details
1.3

What to Ask Before Buying "Agentic"

Three questions separate a considered architecture from a repositioned chatbot, and none require technical depth to ask. First: what actions can this system take without a human, and against which endpoints? A vendor unable to enumerate the action set has not bounded it. Second: when the agent is uncertain, what does it do? Systems that always produce an answer are systems that cannot signal doubt, and a compliance system that never says "unresolved" is not confident, it is uninstrumented — the discipline behind four-state readiness, where missing evidence resolves to UNCLEAR rather than silently to a pass. Third: what does the record show afterwards? If the agent's reasoning, sources and human interventions are not reconstructable months later, the oversight the EU AI Act requires cannot be evidenced even where it occurred.

Get the details
Sources
  1. Gartner Survey Finds Just 15% of IT Application Leaders Are Considering, Piloting, or Deploying Fully Autonomous AI Agents (30 September 2025)
    Other sourceRetrieved August 7, 2026
  2. Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (AI Act)
    EURetrieved August 7, 2026
  3. Regulation (EU) No 952/2013 laying down the Union Customs Code
    EURetrieved August 7, 2026
Current layer Overview