GROUND GROUND
Warehouse Tech Weekly
VOL. 19 2026.09.07

Global Logistics Facility Tech Weekly

Global Warehouse & Logistics Tech Digest

Period covered 31 August – 6 September 2026 Written by Tomo Miyata

This week I chose stories about large third-party logistics providers deciding to own the software that ties machines together.

Kenco is an American third-party logistics provider running around 140 logistics facilities across the United States. It is tripling the automation test facility at its Chattanooga, Tennessee headquarters to roughly 2,790 square metres, opening on 10 September, with more than 30 types of robots on a single floor. The reason for the expansion is that evaluating equipment one machine at a time tells you nothing about what happens once the machines are combined. The first workflow on show links three steps into one: a robot unloads pallets from a trailer, a second moves them deeper into the building, and a third takes the cases off the pallet. Gray Matter software issues the instructions. The facility is run on a deliberately vendor-neutral basis. The work of comparing and verifying what different manufacturers propose is now being done in-house, by the buyer.

In South Korea, CJ Logistics has put two humanoid robots into a live packing process. It buys the hardware from Robotis and the robotic hands from Aidin Robotics. The control software — what it calls a robot foundation model (RFM) — it develops itself, using its own logistics work data. GXO, the world’s largest pure-play contract logistics provider, split its technology organisation in two on 3 September, giving the job of keeping enterprise systems running and the job of putting AI and robotics onto the floor to different executives.

What the three have in common is that none of it is about which machines to buy. Each has decided to hold, as an in-house capability rather than something to outsource, the question of how machines connect to other machines and to the people working alongside them. The buyers have dropped the assumption that manufacturers will handle integration for them — that is how I read it.

There is enough evidence now to justify that view. Nomagic of Poland has made the shoebox — awkward precisely because the lid merely rests on the base — handleable across about 98 per cent of SKUs at up to 450 boxes an hour. Individual hard problems get solved, one at a time. Dexory of the United Kingdom, meanwhile, argues that what a WMS or an ERP holds is a record of transactions, not what is happening on the floor right now. A pallet is moved after it has been scanned; a location is only partly occupied. The more machines there are, the wider the gap between the record and the floor becomes. Equipment can be bought. The state of the operation that the equipment depends on does not come with it.

From here the picture is not comfortable for Japanese operators. Kenco, CJ Logistics and GXO all have the scale to build this themselves. Few companies in Japan do. If that is so, what will separate them is whether they can buy, as a product, the software that tells the machines what order to work in and where to send things. GROUND built GWES with that instruction-issuing function at the centre of the product, rather than joining third-party packages together after the fact. The more varied the items, and the more workflows running through the same building, the more that difference shows — which describes most logistics facilities in Japan.

GROUND Inc. — Founder & CEO Tomo Miyata

This Week’s News

Five stories
01 Lead story
Interior of a large logistics facility filled with automation equipment
Kenco is expanding its automation test facility from about 930 to about 2,790 square metres and putting more than 30 types of robots on one floor. Photo: Unsplash

Logistics OS & OrchestrationUnited States

Kenco triples its test facility to evaluate more than 30 types of robots in combination

Kenco is expanding the automation test facility, its Innovation Lab, at its Chattanooga, Tennessee headquarters from about 930 to about 2,790 square metres. A ribbon-cutting ceremony is scheduled for 10 September. The expanded facility will house at least 30 types of robots across more than a dozen categories of technology. Kenco is a third-party logistics provider that runs distribution operations for its customers, operating around 140 logistics facilities in the United States and also leasing and maintaining the forklifts used at those sites. The lab first opened in 2015.

The purpose of the expansion is to stop evaluating point solutions one at a time and instead test multiple robots working together as a system. “In a 10,000-square-foot space, you can really gobble it up really quickly with point solutions and technologies,” said Ainsley Williams, Vice President of Automation and Innovation. “We realized that if we had a bigger footprint, we could test more technologies as they work together in a system.” A single palletising or depalletising robot would have consumed virtually the entire previous lab on its own. The additional space also allows vertical testing, such as cycle-counting towers that travel up and down the aisles. That reflects an industry shift toward high-bay automation as floor space grows scarce.

The lab’s first major showcase is an end-to-end workflow: a robot unloads pallets directly from trailers at the dock, a second transports them deeper into the building, and a third depalletises the cases. All of it is orchestrated by Gray Matter software, with AutoStore and GrayOrange’s HiCLiME among the platforms on the floor. The project Williams singled out as particularly high-impact is random mixed-SKU pallet building. Current robotic palletising solutions require cases to arrive in a precise sequence — workable alongside automated storage and retrieval systems, but impractical in most 3PL environments. Removing that constraint, he said, would open pallet-build automation to facilities of more than 70,000 square metres. Williams stressed that the lab is deliberately vendor-neutral. “A failed test in a lab is still a successful test in a lab because you wanted to watch it fail. That’s kind of the purpose of a lab — to push it to failure. That’s not the purpose of production.”

Source: FreightWaves / 2026.09.03

A white humanoid robot standing against a black background
Photo: Unsplash
02

Humanoid DeploymentSouth Korea

CJ Logistics puts two humanoids on a live packing process and develops the control software in-house

CJ Logistics has deployed two humanoid robots at the distribution centre serving the cosmetics retailer Olive Young in Yangji, Yongin, Gyeonggi Province. The robots place cushioning material inside boxes during packaging. The company says this is its first use of such robots in an actual logistics process. It follows field testing at its Gunpo fulfilment centre the previous year, now moved into daily operations. Data gathered from the real work will be used as training material to improve the robots’ accuracy and responsiveness.

The company points to two differences from fixed automation equipment designed for standardised tasks: humanoids can work within spaces built for people, and they can potentially shift between different jobs as products and orders change. Control uses CJ Logistics’ own technology, with a robot foundation model (RFM) developed from logistics work data. The RFM serves as the robot’s brain, combining information from vision and sensors with simulated data so the robot can recognise its surroundings and determine how to act with unfamiliar products and work environments.

CJ Logistics plans to extend humanoid operations to picking, sorting, inspection and packaging, with the eventual goal of a single robot performing multiple logistics processes. On the technology side it works with Robotis for hardware, Aidin Robotics for robot-hand technology and RealWorld AI for the RFM. The company is also participating in government-backed projects, including efforts to develop humanlike robotic hands and broader systems for autonomous logistics operations.

Source: The Korea Times / 2026.09.03

Monitoring consoles lined up in a control room under blue lighting
Photo: Unsplash
03

Organisation & Technology StrategyUnited States

GXO splits its technology organisation into Foundation and Acceleration under a new CIO and its existing CTO

GXO Logistics, the world’s largest pure-play contract logistics provider, announced on 3 September an expansion of its technology organisation. Balaji Rangaswamy joins in the newly created role of Chief Information Officer, reporting to Chief Executive Patrick Kelleher. The organisation will be structured around two complementary priorities, Foundation and Acceleration.

Foundation, led by Rangaswamy as CIO, covers the enterprise systems, architecture and core technology capabilities that enable scale. That includes information strategy, information risk, data governance, cybersecurity, digital resilience and service management. Chief Technology Officer Nizar Trigui takes the Acceleration mandate, covering the expansion of GXO IQ, agentic AI, physical AI, robotics and other emerging technologies with the potential to transform supply chain operations.

“As the company and our technology ambitions have grown, we’ve sharpened our focus on two critical priorities: building a scalable, resilient technology foundation and accelerating innovation. These mandates are deeply connected,” Kelleher said, adding that Trigui will lead the innovation agenda while Rangaswamy’s experience in large-scale technology transformation will strengthen the foundation.

Source: GlobeNewswire / 2026.09.03

An operator checking an automated conveyor system inside a logistics facility
Photo: Unsplash
04

CommentaryData FoundationsUnited Kingdom

Dexory: the gap between the record and the floor is capping what automation can deliver

A contributed piece by Oana Jinga, Co-Founder and Chief Commercial & Product Officer at Dexory, which builds robots that move through logistics facilities scanning inventory and the software that presents the results. Despite autonomous mobile robots, automated storage and retrieval systems, AI-powered warehouse management systems and advanced picking technologies, she writes, many facilities still rely on outdated information when making operational decisions. Inventory is constantly moving: pallets are relocated, storage locations fill and empty, and congestion builds in busy areas. By the time systems register these changes, operations have already moved on.

A WMS, an ERP and automation equipment generate large volumes of data, but much of it reflects transactions rather than real-world conditions. A pallet may be moved after it is scanned. A location may be partially occupied. Inventory may be misplaced, or the record itself may be incorrect. As automation increases, the gap between what systems believe is happening and what is actually occurring becomes more critical. Automated systems execute against the inputs they are given, so where data is outdated they can only optimise against an imperfect view of reality.

The same applies to AI, she writes. Demand forecasting, slotting and labour planning all depend on data quality; if inventory records or facility conditions do not reflect reality, even advanced models will produce flawed recommendations. The effectiveness of AI is directly tied to the accuracy of its inputs. On measurement, she argues that traditional metrics such as pick rates, inventory accuracy and throughput need to be joined by a continuously updated view of overall facility health as conditions change through the day. Rather than relying on periodic audits or manual investigations, managers should be able to see where storage capacity is becoming constrained, where bottlenecks recur and how space is being used, and to spot exceptions before they disrupt outbound operations.

Source: Logistics Business / 2026.09.02

An industrial production line lined with machinery
Photo: Unsplash
05

Physical AIRobotic PickingPoland

Nomagic handles about 98% of shoebox SKUs at up to 450 boxes per hour

Nomagic, a Polish company building robots that use AI-based perception to grasp objects, has shown a picking system built specifically for shoeboxes. According to Oscar Cutts, Business Development Manager at the company, shoeboxes make up around 20 per cent of all fashion e-commerce merchandise. Automation has become reliable at moving pallets, cartons and standardised totes, but the shoebox had remained resistant.

A shoebox appears to be an ideal object for a robot. It has defined edges, fairly consistent dimensions and is less deformable than apparel. Unlike sealed cartons, however, most shoeboxes consist of two separate pieces: a base and a loose-fitting lid. Slight variations in how the lid overlaps the base, in the orientation of the box or in the friction between the two can cause the lid to shift or separate during handling. Facilities process hundreds of shoebox designs, sizes and materials, with inventory changing continuously through the day. An elastic band would let a standard suction gripper lift the box from any side, but following trials and surveys, leading shoe manufacturers and retailers found banding detrimental to the user experience and the brand, and now specify that bands must not be used by their distribution partners.

Nomagic’s Shoebox Picker combines AI-based perception with specialised end-of-arm tooling that evaluates each box and adjusts its grasp to the box’s dimensions, lid configuration and orientation. AI vision systems generate three-dimensional representations of each object, allowing the robot to assess size and orientation before planning a grasp; machine learning models then determine the gripping strategy, while tactile sensing verifies whether the grasp succeeded. If conditions change unexpectedly, the robot reassesses and adapts rather than failing. The company says the system handles approximately 98% of shoebox SKUs at picking rates of up to 450 shoeboxes per hour.

Source: Logistics Business / 2026.09.03

← Back to Warehouse Tech Weekly Photos: Unsplash  © 2026 GROUND Inc.