A package crossing into the United States used to depend on a person to review it, classify and catch any errors, slowing the process with days-long delays at the border. Sorting more packages faster is a warehouse problem. Reviewing them accurately is a judgment problem, and judgment has historically required more people as volume grows.
UPS has shown how quickly that can change. The United States eliminated its de minimis exemption globally effective Aug. 29, 2025, ending duty-free treatment for shipments valued at $800 or less, under an executive order President Donald Trump signed last July. The change triggered what UPS CEO Carol Tomé called a “ten-fold surge in daily customs entries,” she said on the company’s third-quarter 2025 earnings call.
In March 2025, 13,000 packages arrived daily that required dutiable clearance, and UPS cleared about 21% of them without manual intervention, Tomé said. By September, 112,000 packages a day required clearance, and the company cleared 90% of them the same way, Supply Chain Dive reported. The share needing no human review climbed from one in five packages to nine in ten, even as daily volume grew nearly ninefold.
How Agentic AI Now Runs Nearly the Entire UPS Customs Process
UPS built that jump on agentic AI, software that carries out a multistep task on its own rather than answering a single question when asked. “We responded swiftly, upgrading our shipping systems to capture the expanded data requirements mandated by U.S. Customs and Border Protection,” Tomé said on the call. “To manage the increased volume and complexity, we enhanced our customs brokerage capabilities by integrating agentic AI. This advanced technology streamlined formal entry processes.”
The system handles classification, documentation and error checking as one continuous process rather than three separate steps a person used to complete in sequence. UPS said 97% of its shipments now clear customs on their first day of entry, and it credited automated product classification, digital trade documentation and more predictable landed costs at checkout, according to a June company announcement.
UPS is applying the same logic to how it plans and monitors its entire physical network. The company is building a digital twin—a virtual replica connecting its facilities, every transportation mode and end-to-end package flows—that refreshes every 10 minutes rather than relying on periodic manual reports. Network analyses that once took engineering teams months can now run in a single afternoon, and forecast accuracy has improved by as much as 40%, UPS said.
Where UPS Goes Further Than Most Enterprise AI
Most enterprise AI deployments still function as an assistant sitting beside a human worker. In financial crime compliance, for instance, agents typically can investigate routine monitoring and screening alerts and resolve clear false positives, while human investigators take the ambiguous cases the regulatory judgment calls, Flagright Co-Founder and Chief Technology Officer Madhu G. Nadig wrote in the PYMNTS eBook, “Building the Agent-Ready Payments Enterprise.” The UPS customs system runs closer to full workflow itself, with a person stepping in only for the roughly 10% of cases the software cannot resolve alone.
Scaling an automated clearance rate from 21% to 90% while absorbing nearly nine times the volume is the kind of result other high-volume, rules-based operations will be measured against, from insurance claims processing to trade compliance elsewhere in the supply chain.
The same visibility push extends beyond the border. UPS has expanded RFID tracking across its U.S. delivery network, installing sensors in delivery vehicles and facilities and tagging shipments originating from more than 5,500 UPS Store locations, The Next Web reported. Package misloads have fallen nearly 70% since the rollout began three years ago. Feeding that RFID data into AI systems gives UPS near real-time visibility into individual packages as they move through the network, and it gives the customs software clearer data on which to act.
