Managing returns is one of the most complex operations in fashion e-commerce logistics, with items poorly packed and in varying conditions. Today, Sereact, Zalando, and CEVA Logistics launched automated returns using robots and AI in fulfillment centers in Greven, Germany, and Świebodzin, Poland.
“One of the toughest challenges is returns handling, due to its extreme operational nature and scale, unpredictable volumes, manual quality checks, seasonality, value decay, and a disproportionately high cost per unit,” stated Chris Walton, senior vice president for contract logistics at CEVA Logistics.
“Our robotics-as-a-service model classifies items faster, reduces manual touchpoints, and delivers consistent, predictable quality on a large scale,” he added. “For our customers, this transforms returns from a cost driver into a controlled, high‑performance service.”
Marseille, France-based CEVA Logistics offers a range of customized logistics systems, as well as multimodal transport in more than 170 countries. The third-party logistics provider (3PL) has about 110,000 employees at more than 1,700 facilities. With 2025 revenue of $18.3 billion (U.S.), CEVA is part of the CMA CGM Group.
Sereact Cortex enables CEVA to sort novel, returned items
Returned items often arrive poorly packed and in varying condition. Examining, sorting, and reconditioning products for reintegration into outbound inventory is typically a labor-intensive process that has been beyond conventional robotic pre-programming, noted the partners.
“Returns are the hardest manipulation problem in fashion logistics,” noted Dr. Ralf Gulde, co-founder and CEO of Sereact. “If a system only works on articles it has seen before, it does not work in a returns operation at all. Running this in production with CEVA at Zalando volumes is the strongest proof point we have.”
Sereact has developed Cortex, physical AI software that enables robots to grasp, identify, and handle unfamiliar items. The Stuttgart, Germany-based company has raised $116 in Series B funding, with Zalando as an investor. It counts IKEA, DHL, Rohlik, Arvato, BMW, Mercedes-Benz, and Daimler Truck among its customers and conducts research and deployment at the same time.
“Generalizing picking will require a data set that includes system dynamics of the real world,” Gulde told Automated Warehouse. “Shortly after ChatGPT came out, we released PickGPT. We realized that the technology behind LLMs were transformative.”
“Instead of building in the lab, we build on the shop floor, aligning with enterprise data collection,” he explained. “We started with the niche of single-arm warehouse pick and place and collected data on 1 billion picks, from groceries to metal parts. Then we went to more complex use cases like kitting for automotive and returns handling. Customers have told us that we have strong credibility in the market.”
The new dual-armed robots use Cortex and have entered live operations to autonomously process returned fashion items, marking the initial phase of a broader initiative to scale returns automation across Europe. Cortex groups previously unseen objects, enabling the system to identify, grasp, and sorting items dynamically without requiring article-specific training or pre-determined SKU sets.
“We’ve learned that the robot has to distinguish between packaging material and trash,” said Gulde. “The craziest item returned was a bag of potatoes.”
The technology is designed to work alongside humans. By taking on highly repetitive handling, the system reduces physical strain on employees, allowing them to upskill into higher-value roles where human expertise and judgement are required, said the companies.
Zalando uses RaaS
Founded in Berlin in 2008, Zalando said it is a leading European technology platform for fashion and lifestyle. The group connects 62 million active customers with more than 7,000 brands across 29 markets. It claimed that its multi-app approach provides a personalized shopping experience, serving different customers with different needs.
“At the heart of our collaboration with Sereact and CEVA is a fundamental belief: True innovation happens when technology and human potential advance together,” said Masood Choudhry, senior vice president of logistics at Zalando. “By solving the most complex challenges in modern fulfilment today, we are actively shaping the future of e-commerce logistics for fashion and lifestyle.”
Zalando said its operations are built on a unique AI-powered data and infrastructure platform. The company has opened its logistics, software, and service capabilities to brands and retailers through ZEOS, Tradebyte, and SCAYLE to help them scale their businesses across borders.
“The system can grade, fold, and repack goods in an appealing way,” he said. “One of Zalando’s biggest problems was that it got 500 to 600 units per hour. Humans can handle only 20 to 30, which is costly, while robots can work without breaks. Zalando was willing to take a risk with Sereact in the beginning. They visited our demo and then visited it again two weeks later, and they appreciated the pace of learning.”
One impetus for more intensive returns processing is that the EU has made it illegal to destroy returned goods or send them to landfills.
“Companies can save transport costs and restock quicker, making sure that seasonal goods can be resold,” noted Gulde. “We look at all goods to make sure the quality is high to reach maximum customer satisfaction.”
Sereact and CEVA Logistics deliver automated returns processing to Zalando through a robotics-as-a-service (RaaS) model. The partners said that the deployment is the first operational milestone in their strategic collaboration. They added that they are actively working to scale intelligent automation across their shared European logistics footprint.