In a traditional warehouse, filling an order means a person walks. They take a list, push a cart or carry a basket, and travel up and down long aisles gathering items from shelves that may be spread across a building the size of several football pitches. A great deal of the working day is spent not picking but walking between picks - covering distance, searching for locations, retracing steps. One of the more effective applications of robotics in these buildings attacks exactly that waste, and it does so through a simple inversion: instead of sending the worker to the goods, it brings the goods to the worker.
In a goods-to-person system, mobile robots lift and carry the storage shelves themselves, ferrying them from wherever they sit in the warehouse to a fixed station where a person stands. The worker no longer walks the aisles; they remain at the station while the robots deliver, in sequence, the shelves holding the items needed for the current orders. The person picks the item, the robot returns the shelf and brings the next, and the walking - which was never the valuable part of the job - is absorbed almost entirely by the machines. The human contributes what humans are good at, the dexterous recognition and handling of varied items, and is relieved of what they were bad at and hated, the endless trekking.
The reason this pays off is that it targets the genuinely wasteful part of the task rather than trying to replace the hard part. Picking a specific item from a shelf - recognising it, grasping it correctly, handling the huge variety of shapes and sizes real inventory contains - remains difficult for a robot and easy for a person. Walking a warehouse is easy for a robot and a poor use of a person. Splitting the task along that line lets each side do what it does well: the robots handle the movement and the storage density, the people handle the discrimination and the handling, and the throughput of the whole operation rises without either side being asked to do the thing it is bad at.
As with warehouse automation generally, the case for it is specific rather than universal, and turns on scale and product mix. The economics favour large operations with high order volumes, where the walking really does consume a significant share of labour and the investment in robots and the systems to coordinate them can be spread across enough activity to pay back. A small operation where a person can reach everything in a few steps gains little. Where the conditions fit, goods-to-person systems are a clear example of robotics deployed sensibly - not a humanoid worker meant to replace the person, but a reorganisation of the physical task so that machines carry the shelves and people do the picking, each freed to do the part that suits them.