From one pilot site to a working hub
FedEx and the robotics company Dexterity have expanded their trailer-loading deployment from a single pilot site into full production at FedEx's Hagerstown Hub in Maryland. The companies describe it as the next stage of a partnership built over several years of joint development, testing, and operational validation, not a fresh experiment.
The distinction between a pilot and production matters more than it sounds. A pilot proves a robot can do the job under controlled conditions. Production means the robot is now part of the hub's actual daily throughput, running alongside human loaders on real freight, not a demo running in a corner of the facility.
What the robot actually does
The system pairs Dexterity's Foresight platform, what the company calls a physical artificial intelligence (AI) world model, with a dual-armed robot it calls Mech. Mech uses vision, depth sensing, and touch to make loading decisions in real time: how to position each parcel for space efficiency, how to keep the load stable, and how to keep the pace up as parcels keep arriving. That combination of senses is what lets it handle irregular, unlabeled packages rather than just picking from a fixed, pre-sorted bin.
Why trailer loading, specifically
FedEx moves tens of thousands of trailers a day across its United States (US) network, and loading them is one of the physically toughest jobs in that chain. It calls for sustained strength, endurance, and constant, fast problem-solving as parcels of every size and shape come down the line. FedEx has pointed to loading as one of the areas where its own team members most want additional support, which is the practical reason it is the task getting automated first rather than something further up the value chain.
Kawal Preet, FedEx's Executive Vice President of Planning, Engineering and Transformation, said the expanded collaboration with Dexterity "reflects our commitment to advancing capabilities that improve how we operate today while helping prepare our network for the future." Dexterity founder and chief executive officer (CEO) Samir Menon said Hagerstown "represents an important milestone in demonstrating how physical AI can operate reliably in a high-volume logistics environment."
Still the exception, not the norm
That second quote is a fair description of where the industry actually stands. A June 2025 survey by Talking Logistics and Indago found that only 4% of supply chain operators had deployed robotics beyond a single pilot for trailer or container loading and unloading. Most companies in that survey had not put a loading robot into even one site. Measured against that baseline, moving a deployment from pilot to full production at one hub is a genuine step past where most of the industry has managed to get, even though it is still just one hub out of FedEx's national network.
Part of why so many pilots stall is that operators cannot tell if they are working. Separate Gartner research, cited alongside the Indago survey, found that 55% of supply chain leaders are unclear on the return on their AI investments. That is a broader finding about AI spending generally, not robotics specifically, but it points at the same underlying problem: a lot of companies are being pushed to deploy automation before they have built the measurement tools to know whether the last deployment actually paid off. A multi-year validation process, of the kind FedEx and Dexterity describe here, is one way to avoid finding that out after the robots are already on the floor.
What comes next
FedEx says it is now evaluating where else physical AI could help around its hubs, including trailer assignment, destination planning, maintenance work, and other workforce processes. None of that is committed yet. Hagerstown is the proof point the company will presumably point to as it decides how far, and how fast, to extend automated loading to other hubs.
What it means in practice
For logistics operators watching from the outside, the useful signal here is not the robot itself, it is the timeline. Dexterity and FedEx spent multiple years validating this system before calling it production-ready at even one site, which is a longer runway than most vendor pitches suggest is necessary. Anyone evaluating a warehouse or dock robotics vendor should ask how long that vendor's own reference deployments ran as pilots before being trusted with real freight, and treat a fast pilot-to-production timeline as a reason for more scrutiny, not less.
For supply chain teams under pressure to show automation progress, the 4% figure is the more sobering number. Most peers have not gotten a robot past a single pilot site at all. That gap between AI investment and actual operational deployment is where a lot of current supply chain technology spending is currently stuck, and Hagerstown is one of the few concrete counterexamples on record.
