Warehouse robots did not replace the workers. They replaced the walking.

The simple story about warehouse automation is that robots take jobs. The measured story across a decade of goods-to-person deployments is stranger and more interesting.

Headcount per unit shipped fell, but not as sharply as the vendor projections suggested. What changed decisively was what a shift felt like. When the shelves come to the picker, the picker stops walking, and the walking had been the unplanned rest.

The ergonomic trade

Operators report fewer musculoskeletal injuries associated with carrying and more associated with repetitive reach at a fixed station. Net injury rates in published studies move less than either side expects. The injuries simply change category.

Where the labour went

Into exception handling, maintenance and inbound processing. These roles pay better and are harder to fill, which is why several large operators now run their own technician training programmes rather than recruiting for the skill.

The measure operators use now

Not picks per hour. Retention at ninety days, and the ratio of exception events to total throughput. A system with excellent pick rates and constant exceptions costs more in supervisory attention than it saves in labour.

For anyone specifying a system

Ask the vendor for exception rates at a comparable site, in writing, and ask what proportion required a human to enter the automated area. That second number predicts the staffing model better than throughput does.