GROUND GROUND
Warehouse Tech Weekly
VOL. 29 2026.11.15

Global Logistics Facility Tech Weekly

Global Warehouse & Logistics Tech Digest

Period covered 9 – 15 November 2026 Written by Tomo Miyata

I believe that adding more robots will not raise a facility's overall throughput unless there is software to instruct both the robots and the people. This week I chose stories about that software, and essays on why automation falls short of its planned results. The lead story is DHL Supply Chain.

The lead story is an announcement made on November 9 by DHL Supply Chain, the contract logistics arm of Germany's Deutsche Post DHL Group. The company has deployed more than 9,000 robots in logistics facilities around the world. It is now moving software it calls "AI orchestration" from the trial stage into live facility operations.

DHL describes AI orchestration as a "digital control center" that connects data, employees and automated machinery into one team. In one example the company is testing, the software connects autonomous mobile robots (AMRs), packaging automation and the warehouse management system (WMS). It then changes how work is allocated between people and machines according to the order volume at that moment. In terms of existing facility software, AI orchestration is close to the role of a WES and a WCS. A warehouse execution system (WES) decides in what order incoming orders are broken down into tasks and how they are assigned. A warehouse control system (WCS) tells the machines in the facility which one should move what, and where, next.

I expect robots themselves to become easier to obtain. GXO, the US contract logistics giant, reached a preliminary agreement to deploy up to 1,000 humanoid robots from Apptronik, a humanoid maker based in Texas, over four years. It expects to pay a usage fee rather than buy them, under a robotics-as-a-service (RaaS) contract. Geek+, the Chinese logistics robot maker, announced a plan to run facilities with a small staff at night and on weekends.

If robots become easier to obtain, I believe two things will separate one facility's productivity from another's. The first is whether a single piece of software can tell machines from different makers and people what to do, in what order and where. The second is whether people have drawn the line on how much AI is allowed to decide. DHL's executive said the most successful supply chains will not be fully automated, and the company leaves decisions on unexpected situations to people. GXO and Apptronik likewise plan to set productivity and safety criteria for each task and expand deployment only to the tasks that meet them.

Adding machines without software that issues those instructions leaves goods waiting between processes, and the facility stopping more often. An article in the US logistics outlet Logistics Viewpoints gives an example. If a robot builds 100 pallets an hour but the next stretch-wrapping step can handle only 70, the facility can ship only 70 pallets an hour. The other 30 pile up between the two steps. The founder of Thomax, an Australian WMS company, points out that many legacy WMS release instructions in batches every 15 minutes or every hour. Some robots need their next instruction within milliseconds, so if it does not arrive in time, they sit idle.

GROUND's product, GWES, is built around software that instructs both machines from different makers and people. We did not design it as a bolt-on to other vendors' software. Japan's logistics facilities face severe labor shortages and handle a wide mix of goods in a single facility. For that reason I believe that having software that instructs both machines and people will determine the productivity gap between facilities.

GROUND Inc. — Founder & CEO Tomo Miyata

This Week's News

Five stories
01 Lead story
Illustration of a logistics facility
Equipment inside a logistics facility (illustrative). Photo: Unsplash

DeploymentAI orchestrationUnited States

DHL Supply Chain — Moves AI software that allocates work to people, robots and systems into live facility operations

DHL Supply Chain, the contract logistics arm of Germany's Deutsche Post DHL Group, said on November 9 that it is taking its digitalization strategy to the next stage. Tim Tetzlaff, who leads the company's digital transformation globally, outlined the plan at the Plug and Play Silicon Valley Summit. There are two pillars. One is to deploy "AI orchestration," which has completed its trial, in live facility operations. The other is to test AI-powered robots that change how they act as conditions change.

According to the company, AI orchestration is a software layer that acts as a control center, connecting data, AI agents, robots, employees and other machinery into one team. It lets site leaders respond faster to swings in demand and reassign people and equipment on the spot. In one example the company is testing, the software connects autonomous mobile robots (AMRs), packaging automation and the warehouse management system (WMS). It then changes how work is allocated between people and machines according to the order volume at that moment. The new AI-powered robots do not only follow predefined routes or instructions; they work toward a goal and choose a different course of action when their surroundings differ from the plan.

The company has deployed more than 9,000 robots across its global facilities. Robotic picks have passed 1.5 billion. Unloading with Stretch, the case-unloading robot from US robot maker Boston Dynamics, has reached 10 million cases. Analytics and AI solutions have been deployed more than 1,400 times, including more than 500 deployments in 2026. More than 95% of the logistics facilities the company operates worldwide have advanced digitalization solutions. Use of its conversational AI agents has passed 2.5 million minutes in total.

The company aims to grow revenue by 50% by 2030, and says removing people from operations is not the goal. Robots take on physically demanding and repetitive work, and AI searches large volumes of data for patterns and actions to take. People handle the judgment calls in unexpected situations. "The most successful supply chains will not be fully automated," Tetzlaff said.

Source DHL (press release) / 2026.11.09

Illustration of a logistics facility
Photo: Unsplash
02

Humanoid robotsRaaSUnited States

GXO — Plans to deploy up to 1,000 Apollo humanoids from Apptronik over four years, under a usage-fee contract

GXO, the US contract logistics giant, said on November 12 that it has signed a letter of intent with Apptronik, a humanoid robot maker based in Texas, to deploy the robots at scale. From 2027, Apptronik's Apollo humanoid will be tested in live work at GXO logistics facilities. The tasks are picking, case and tote movement, packing and sortation. Picking will be tested both where robots bring shelves of goods to a worker and where workers walk to the shelves.

Each task will be assessed against productivity, safety and operational criteria agreed by both companies, which will also set the order of deployment. If the criteria are met, the companies intend to roll out up to 1,000 Apollo robots globally over the next four years, starting in North America in late 2027. The contract is expected to take a robotics-as-a-service (RaaS) form, in which GXO pays a usage fee instead of buying the robots.

The two companies partnered in 2024, and this year completed a pilot in the US on autonomous picking of small consumer goods. At Robot Park, Apptronik's humanoid training facility, they had the robots perform more than 50 different logistics tasks to collect data and evaluate autonomous operation.

Source MarketScreener (S&P Capital IQ) / 2026.11.12

Illustration of a logistics facility
Photo: Unsplash
03

Autonomous operationWESChina

Geek+ — Unveils a plan to run facilities with a small staff at night and on weekends, built around software combining robot management and a WES

Geek+, the Chinese logistics robot maker listed in Hong Kong, announced on November 10 its "24/7 Warehouse Autonomy Plan" to run logistics facilities around the clock with a small staff. The company says it has ranked first worldwide in autonomous mobile robot (AMR) market share for seven consecutive years, and it has deployments in more than 40 countries and regions.

At the center of the plan is the company's "Warehouse Brain" software. It consists of three parts: a robot management system (RMS), a warehouse execution system (WES) that decides in what order orders are broken into tasks and how they are assigned, and an operations platform (IOP) that analyzes how operations are running. Warehouse Brain decides task assignment, routes, robot traffic control and the allocation of people and equipment across robots, automated equipment and work processes. AI agents spot anomalies and stalled work, investigate causes and propose improvements. The company says it will move the AI agents from proposing improvements toward making decisions themselves, adjusting operations and recovering from faults.

On the robot side, the company applies AI to tasks with items of varying shape and material, and to handling exceptions. People have handled this work until now. The company names three priorities: increasing the range of tasks robots can take on, strengthening the software loop from detection to decision, execution and learning, and adding customers through its deployment and service teams in each country.

Source Morningstar (via ACCESS Newswire) / 2026.11.10

An operator checking equipment inside a logistics facility
Photo: Unsplash
04

EssayReturn on investmentUS / Europe

Logistics Viewpoints — Even when robots perform to spec, upstream and downstream steps can undo the return on investment, an essay argues

Jim Frazer of the US logistics outlet Logistics Viewpoints published an article on the return on automation investment on November 12. A machine's rated capacity is not the same as the capacity a logistics facility actually achieves, he writes. The gap comes from product presentation, packaging, pallet quality, data accuracy, software integration, how people handle exceptions, and whether the next process can accept the output.

The article gives three examples. At XPO Logistics' fresh-goods facility in Nieuwegein, the Netherlands, the criteria for inspecting pallets before they enter an automated crane storage system did not match what the equipment actually needed. After the facility reviewed the inspection process with CHEP, the pallet rental company, and trained employees, both rejected pallets and the use of support pallets fell by about 50%. Staples, the US office supplies company, implemented a WMS from Manhattan Associates across nine US business-to-business fulfillment facilities within 18 months. The article uses this example to show that automated equipment cannot move a pallet unless the software correctly knows what it contains, where it should go and which rules apply to exceptions. Del Papa Distributing, a beverage wholesaler in Texas, linked robotic pallet building through stretch wrapping without a person stepping in.

The article notes that if a robot builds 100 pallets an hour but the next wrapping step can handle only 70, the facility's capacity is 70 pallets an hour. A system designed for 1,000 pallet movements a day may achieve only 900 once material shortages, planned downtime, imbalance between processes and recovery from exceptions are taken into account. The investment case should therefore include how often the following occur: pallet rejections, unscannable labels, unstable loads, task failures, integration outages and manual interventions.

Source Logistics Viewpoints / 2026.11.12

Illustration of a logistics facility
Photo: Unsplash
05

WMSException handlingAustralia

Thomax — Decide before go-live how operations will continue when automation stops, says the founder of an Australian WMS company

Thomas Jackson, founder of the Australian WMS company Thomax, described where automation projects run into trouble in an article published by eCommerceNews Australia on November 13. Robots are strong at repetition and weak at the unexpected. A cracked tote can stop a robotic arm from docking it properly. A person finding an aisle blocked simply takes another route, but a robot cannot respond unless it has been trained for that situation. As smaller systems and robotics-as-a-service (RaaS) contracts spread and more companies start automating, more of them are running into these problems, he says.

He gives the example of a facility handling appliances and homewares. Toasters and laptops fit in a robotic tote, but a 60-inch TV does not. When a games console and a TV are in the same order, the order has to be split between the automated and manual areas and brought back together before shipping. Many legacy WMS also release work instructions in batches every 15 minutes or every hour. Some robots, by contrast, need their next instruction within milliseconds, and sit idle if it does not arrive in time.

Jackson adds that tests in digital twins, which recreate equipment in a virtual space, tend to assume that barcodes always scan and conveyors never jam. He lists five things to check before signing: whether the provider has done this kind of deployment before, how much will be configuration versus new code, whether the system responds in real time at peak, how it handles orders spanning automated and manual areas, and how operations will continue when the automation is offline.

Source eCommerceNews Australia / 2026.11.13

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