UPS Deploys AI to Tackle Return Fraud at Scale

Returns have become one of the most expensive and operationally complex parts of modern retail logistics. The National Retail Federation estimates that consumers will return $849.9 billion worth of merchandise in 2025, representing roughly 16% of total retail sales. Within that volume, about 9% of returns are fraudulent, involving swapped items, counterfeit goods, or mismatched products.

UPS is now pushing AI into that pressure point. Through Happy Returns, the reverse-logistics company it acquired in 2023, UPS has begun deploying an AI system designed to identify potentially fraudulent returns before refunds are finalized. The initiative reflects a broader shift in logistics AI toward systems that filter risk in high-volume, high-ambiguity workflows, rather than focusing solely on optimization.

Returns as a failure point in retail logistics

Unlike outbound fulfillment, returns lack standardization. Items often arrive without original packaging, through decentralized drop-off points, and with limited documentation. In many cases, refunds are initiated before physical inspection occurs, creating both financial exposure and operational strain.

Happy Returns operates roughly 8,000 “return bars” across the U.S., allowing consumers to drop off items without boxes or shipping labels. While the model lowers friction for shoppers and reduces logistics costs for retailers, it also concentrates fraud risk at scale.

To address that risk, Happy Returns began piloting an AI tool called Return Vision during the 2025 holiday season. The system flags potentially fraudulent returns based on behavioral signals and image comparisons, routing them to human auditors for review. Flagged returns are processed at hubs in California, Pennsylvania, and Mississippi, where auditors open packages, photograph the contents, and make final determinations.

“Sometimes humans don’t always catch small differentiation points between an item that’s been returned and the item that was purchased,” David Sobie, CEO of Happy Returns, told Reuters.

For retailers, the financial impact is immediate. “Not getting back the real items is a double whammy. It’s hundreds of thousands of dollars for us alone per year,” said Jim Green, director of logistics and fulfillment at Everlane, which is participating in the pilot.

How AI is being applied, and where it stops

Return Vision relies on computer vision to compare images of returned goods against product photos and metadata associated with the original transaction. The system is intentionally conservative. According to Reuters, fewer than 1% of returns are flagged as having a high probability of fraud. Of those flagged cases, about 10% are ultimately confirmed as fraudulent, with an average fraud value of approximately $261.

“If you’re returning a pair of $300 boots and you show up with a pair of dirty old sneakers, that should be caught immediately,” Green said. “What Return Vision does is add an extra layer of protection for some of the not-so-obvious cases.”

The system does not attempt to detect every form of return abuse. Practices such as wardrobing, where customers wear items and return them without visible damage, remain difficult to identify through image comparison alone. Such cases are not yet addressed by the tool.

Happy Returns’ chief operating officer Juan Hernandez-Campos described the approach as adaptive. “Bad actors adapt. We need to adapt too,” he said.

AI as operational infrastructure at UPS

The return-fraud deployment sits alongside broader efforts by UPS to embed AI and automation across its operations. The company has publicly outlined plans to expand robot-powered sorting, automated bagging, and AI-assisted facilities, as well as the use of digital twins to model and optimize network performance.

UPS has also deployed AI-driven tools for customer service automation, including systems that assist with shipment tracking and service inquiries, and internal platforms such as Languages Across Logistics, which uses machine learning to support multilingual communication and onboarding across facilities.

These initiatives show how UPS is applying AI as operational infrastructure, from fraud detection to customer service and facility automation. UPS CEO Carol B. Tomé has described automation as central to the company’s long-term strategy, pointing to future operations “powered by robots and humanoids and automated sorting and automated bagging and automated label application.”