AI Resilience Report forCustoms Brokers

Customs brokerage is one of the trade jobs where AI is moving fastest, mostly because so much of the work involves paperwork, codes, and rules that machines can learn to read. Trade-compliance vendors now offer machine-learning tools that suggest Harmonized System codes directly from product descriptions — Thomson Reuters says its ONESOURCE classifier targets 95% accuracy and can cut manual research time by 50% ^[1]. Governments are moving in the same direction: Baker McKenzie reports that customs authorities across Africa are embedding AI into cargo profiling, valuation, and predictive risk analytics, with automated systems increasingly performing validation and administrative checks ^[2].

Brokers themselves are adopting private AI too — SupplyChainBrain describes how R.L. Jones Customhouse Brokers used a private LLM to replace manual ACE downloads and per-customer Excel workbooks, letting analysts ask natural-language questions and get refund calculations "in seconds instead of hours" ^[3]. Importantly, this looks more like augmentation than replacement.

At the 2026 NCBFAA Annual Conference, brokers stressed that "technology supports compliance—it doesn't replace it," with AI focused on validation, data quality, and flagging inconsistencies while licensed brokers keep making the final calls ^[4].

Adoption should stay fast for repetitive tasks — classification, duty math, and shipment tracking — because commercial tools are already cheap compared with skilled broker labor, and tariff volatility is pushing importers to demand faster answers. But there are real brakes. The World Customs Organization's BACUDA project warns that AI only delivers reliable results when robust data governance is already in place, and that many customs members try to jump to AI models before fixing "inaccurate, incomplete or inconsistent" data ^[5].

Legal accountability is another brake: transacting customs business still requires a federal license from U.S. Customs and Border Protection under 19 U.S.C. 1641, and signing entries on behalf of clients using powers of attorney is a human legal responsibility ^[6]. On the labor side, Goldman Sachs projects that "the big story in 2026 in labor will be AI," with entry-level knowledge workers likely to feel the impact first ^[7] — so early-career brokers may see routine data-entry roles shrink, while people who learn trade policy, agency negotiation, and AI-tool oversight will remain in demand. The honest takeaway: AI is quickly handling the busywork, but licensed judgment, relationships with CBP officials, and accountability for the paperwork you sign are still very human skills worth building.