The humanoid robot is often presented as a general-purpose machine that will move easily from factory floors to kitchens. Its first sustained commercial work is more likely to be narrower. Warehouses and factories contain repetitive material-handling jobs, mapped spaces, standardized containers, and engineering teams that can redesign a workflow around a machine. Homes contain stairs, pets, clutter, children, fragile objects, and tasks whose success is difficult to specify. The human shape may be general; early deployments are not.
Warehouses make the problem smaller
Industrial buildings were designed around human reach, carts, shelves, doors, and workstations. A bipedal robot with two arms can potentially use that infrastructure without rebuilding an entire site around fixed automation. This is the core practical argument for a humanoid form: not that legs and hands are always the most efficient mechanism, but that they fit environments already built for people.
Those environments also make autonomy easier to constrain. A warehouse operator can define pickup and drop-off points, restrict operating zones, standardize package dimensions, impose speed limits, and keep a human escalation path. Performance can be measured in moves per hour, error rates, uptime, and avoided ergonomic strain. A household request such as “tidy the living room” hides far more ambiguity than moving a tote from a robot to a conveyor.
Commercial trials are deliberately narrow
In June 2024, GXO and Agility Robotics announced a multi-year agreement following a 2023 proof of concept. GXO said it was “deploying fully operational Digit humanoids into a live warehouse environment” at a Spanx facility, using a robots-as-a-service arrangement (Agility Robotics). Digit's assigned workflow moved totes between autonomous mobile robots and conveyors. The task exercised walking, grasping, and coordination, but it did not ask the robot to improvise across the whole building.
Amazon likewise announced that it would test Digit for moving empty totes in its operations. Its 2023 facilities summary described the bipedal robot as suited to spaces and corners where moving, grasping, and handling objects could be useful (Amazon). The empty-tote detail matters: companies are starting with objects whose weight, shape, and consequences of mishandling are relatively predictable.
Manufacturing trials point in the same direction. In 2024, BMW tested Figure 02 for several weeks at its Spartanburg plant, where it inserted sheet-metal parts into fixtures for chassis assembly (BMW Group). Figure later reported an 11-month deployment contributing to production of 30,000 vehicles, although that performance account comes from the robot supplier and should be read as a company claim rather than independent validation (Figure).
The business case depends on reliability
Warehouse economics are unforgiving. A humanoid does not compete with a research demonstration; it competes with a person, a conveyor, an autonomous mobile robot, or a conventional arm that may perform one job faster and more reliably. A flexible robot becomes attractive when work changes often, floor space is constrained, or fixed automation is too expensive. It is less compelling when a simple machine can repeat a stable motion continuously.
Robots-as-a-service contracts shift some capital cost and maintenance risk to the vendor, but customers still bear integration costs. The machine must connect to warehouse-management software, charging routines, safety systems, exception queues, and maintenance schedules. Useful autonomy includes knowing when to stop and request help. A robot that completes most moves but regularly requires a technician may create more operational friction than it removes.
Safety is a system property
A human-shaped robot working near people introduces risks from its mass, momentum, battery, end effectors, carried objects, and unexpected motion. OSHA notes that many robot accidents occur during non-routine conditions such as maintenance, testing, setup, and adjustment (OSHA). Its technical guidance calls for an application-specific hazard analysis and risk assessment, including the employer and workers, for collaborative systems (OSHA Technical Manual). Safety cannot be established by a promotional video showing smooth motion in an empty aisle.
Deployment also changes work rather than simply deleting it. People may move away from repetitive carrying while taking on fleet supervision, exception handling, charging, inspection, and repair. Whether that is an improvement depends on training, job design, staffing, and whether productivity gains are shared. Measuring injury exposure and work intensity is as important as counting automated moves.
Homes remain the harder frontier
Domestic robots must be inexpensive, quiet, energy-efficient, socially acceptable, and safe around people who have received no workplace training. They must manipulate unfamiliar objects and understand preferences that vary by household. A warehouse can mark a no-go zone; a toddler will not respect one.
Industrial deployments therefore should not be mistaken for general intelligence in a mechanical body. They are valuable because they convert an open-ended robotics problem into controlled, auditable work. If humanoids earn a place in warehouses, it will be through reliable execution of a growing set of bounded tasks. The route to the home, if it exists, runs through thousands of uneventful shifts at work.