Otto Group Deploys 25 NVIDIA-Powered Robots to Sort Returns in Poland

Otto Group and Sereact will roll out up to 25 dual-arm robots running Cortex, a physical AI model that can handle unfamiliar items without prior training.

Key Takeaways

  • Otto Group and Sereact announced on October 28, 2026 a plan to deploy up to 25 dual-arm robots to sort returns at a Polish logistics center.
  • The robots run on Cortex, Sereact's physical AI model, which can handle items it has never seen before without product-specific training.
  • The rollout happens in two phases: 10 robots first, then 15 more, operated by Sereact under a service model with human quality oversight.
  • The project is part of Otto Group's broader €350 million medium-term investment in logistics technology and AI.

Otto Group Deploys 25 NVIDIA-Powered Robots to Sort Returns in Poland

Otto Group will deploy up to 25 AI-controlled robots at a logistics center in Poland to grasp, identify and sort returned items, in a project with German robotics software firm Sereact announced October 28.

The rollout happens in two phases: ten dual-arm robots first, followed by fifteen more once the initial batch proves out. The robots run on Cortex, Sereact's physical AI model, which the company says can recognize and manipulate items it has never encountered before — no fixed product catalog, no item-by-item training. "A robot that only knows yesterday's catalog is of little use in a returns center," Sereact's Dr. Ralf Gulde said, according to the Sereact announcement. The stack runs on NVIDIA Isaac Sim, Isaac Lab and Cosmos tools, powered by NVIDIA Blackwell GPUs.

Returns are the ugly side of e-commerce logistics

Returns processing is one of warehouse automation's hardest unsolved problems because, unlike inbound stock, returned items arrive damaged, mislabeled, repackaged or simply unidentifiable at scale. Most warehouse robots today are trained against a known SKU catalog; Otto Group's bet is that a foundation model trained to generalize across unknown objects can finally make returns sorting profitable to automate rather than a pure labor cost.

Sereact will own and operate the robots under a service contract rather than selling hardware outright, with Otto Group employees supervising, handling exceptions and signing off on final quality decisions. That human-in-the-loop structure mirrors how most current warehouse robot deployments — including Amazon's million-plus robot fleet and GXO's humanoid pilots with Agility Robotics — still rely on people for edge cases the AI can't resolve.

Part of a bigger spending wave

The Poland project is one piece of Otto Group's committed €350 million medium-term investment in technology and AI across its logistics operations, according to Otto executive Martin Umland. It also lands amid a broader surge of warehouse robotics spending: Amazon crossed one million deployed warehouse robots in October, and rivals including Ocado and DHL have all expanded automated sorting trials this year.

Sereact has not disclosed pricing or a timeline for the second phase of ten additional robots. If the Poland pilot hits Otto Group's efficiency targets, the company has signaled it plans to extend Cortex-based sorting to other logistics centers in its network.

Category: Industrial Robots

Tags: Physical AI Nvidia Otto Group Sereact warehouse robots returns processing

Related Articles

Frequently Asked Questions

How many robots is Otto Group deploying?

Up to 25 dual-arm robots in two phases — an initial 10 followed by 15 more — at an Otto Group logistics center in Poland.

What is Sereact's Cortex model?

Cortex is Sereact's physical AI model that can recognize and handle items it has never previously been trained on, removing the need for a fixed product catalog in robotic sorting.

What hardware powers the project?

The system runs on NVIDIA Isaac Sim, Isaac Lab and Cosmos tools powered by NVIDIA Blackwell GPUs.

Who operates the robots day to day?

Sereact operates and maintains the robots under a service contract, while Otto Group employees supervise the process, handle exceptions and make final quality decisions.