

From shoebox to trench coat: the full fashion assortment, robot-ready.

Long garments, from coats and maxi dresses to anything on a hanger, have traditionally caused problems in warehouse automation. Conventional systems either couldn’t pick them at all, or ran into issues while handling them: the items dragged, snagged on equipment and couldn’t be placed properly. Nomagic’s new long-item automation changes that. It combines AI that recognises these items and works out how to pick and manoeuvre them with a newly designed gripper built for length. It’s already live in production at Zalando.
Polybagged apparel makes up a large share of fashion fulfilment centre volume. It has never been easy to automate: the packaging is reflective, the contents shift, and no two items deform the same way. Nomagic’s robots have handled these challenges reliably across tens of millions of picks in live operations and multiple customer sites.
Yet full warehouse automation in fashion has been held back by a small group of non-standard formats, including shoeboxes, long garments and garments on hangers.
We have been closing that list one item type at a time. First came shoeboxes: our Shoebox Picker made them a routine pick.
The next gap was length. In theory, a robot can pick coats, maxi dresses and garments on hangers the standard way. In practice, they trigger interventions and other problems far more often than regular items. Our new long-item automation closes that gap.
A robot station is built around typical item sizes. A floor-length coat is not typical. Picked with the standard polybag tool and flow, long items caused problems at almost every step:
Each of these meant a stop and a person stepping in. The alternative was a decision not to pick the item at all. Skipping it caused problems of its own: long items often cover other items in the bin, so those couldn’t be picked either.

This innovation combines AI with new hardware: AI that recognises long items and works out how to pick and manoeuvre them, a newly designed gripper the robot picks up on its own, and a new motion path built for length.
When a long item appears in a bin, the system works through four steps:
As a result, long items are handled automatically, the rest of the bin keeps moving, and no manual intervention is needed.
This isn’t a lab prototype. The long-item automation runs today on 13 production robots across 4 client sites and handles around 7,500 long items every week, with no operator stepping in. Handling long items enabled us to perform 15% more robotic picks from the same number of bins.
| 13 robots, 4 sites | Already in production |
| Under 2 months | From business need to full production deployment |
| 83% | Fewer interventions caused by long items |
We went from business need to a solution fully deployed in production at scale in under two months.
That speed is by design. Nomagic’s platform is built to adapt: new AI capabilities, new tools and new motion paths can be rolled out to robots already working in the field. When the market shifts or a client’s assortment brings a new challenge, we respond in weeks, not years.
The hard part is already done. The robot can recognise a long item, switch to the right tool by itself and move a floor-length garment through the cell without touching a thing. Those are the building blocks for handling long items in whatever flow a site needs.
Put it together with the Shoebox Picker, accessories handling and the standard polybag flow, and the picture is clear: the formats that once kept fashion out of reach for robots are falling one by one. Fashion is becoming fully automatable.
Long garments like coats and maxi dresses are a practical manipulation problem for warehouse robots: they’re floppy, catch on equipment, and often block other items in the bin. Nomagic solved it with a self-learning AI system that recognises long items, plans each pick and guides them safely through the station, paired with a highly efficient dedicated gripper that lifts and holds these garments securely.
Deployed in under two months, the solution now runs on 13 robots across 4 customer sites, significantly decreasing interventions. This is Physical AI built for high throughput and high reliability.
Have items your automation still can’t handle? Let’s talk.


