In pilot testing inside an Amazon fulfillment center, a bipedal robot named Digit, built by Agility Robotics, spent its shift doing one narrow task: picking up empty totes and moving them from one conveyor point to another. It did not greet anyone, fold anything, or navigate stairs. It did the single most repetitive part of a warehouse job, over and over, without getting tired.
Warehouses offer almost every condition a robot needs to succeed that a home does not — flat, predictable floors, controlled lighting, structured shelving, and tasks that repeat identically thousands of times a day. A home offers stairs, pets, clutter, and tasks that differ every time, a much harder engineering problem than humanoid-robot marketing tends to admit.
Optimizing for the warehouse case — repetitive, structured, high-volume — produces machines that are economically justified quickly but not obviously transferable to unstructured environments. Symbotic's multibillion-dollar automation partnership with Walmart, retrofitting distribution centers with robotic pallet and case-handling systems, is built on the same logic: narrow the task until the economics work, then scale that narrow task everywhere it applies.
Warehouse and logistics labor is where the near-term effects concentrate, and the pattern looks less like wholesale replacement than task reallocation — robots handling repetitive movement, humans handling exceptions, damaged goods, and judgment calls the robot's training data never covered.
The economics of robotics in logistics are improving specifically because the task is so narrow, which means the companies profiting first — Symbotic, Agility, and the warehouse operators buying from them — are the ones building for that narrowness rather than the ones chasing a general-purpose humanoid that can do everything a person can, the way Figure AI framed its manufacturing trial at BMW's Spartanburg plant as a pilot rather than a full deployment.
The home-robot narrative will likely arrive the way self-driving did — first in the most constrained version of the problem, a single task in a controlled space, rather than as one general leap. Digit moving totes on a flat floor is the honest current state of the art. Folding laundry in someone's actual living room is still, by comparison, a research problem.