Humanoid robots are moving closer to industrial production through deployments that combine large hardware orders with the collection of detailed human-motion data. British robotics company Humanoid plans to place machines in factories operated by German supplier Schaeffler, while South Korea’s RLWRLD is recording workers to teach robots practical movements.
Together, the projects show two sides of Physical AI: integrating machines into existing facilities and creating the training data needed for dexterous work.
They also expose the social questions surrounding worker data, job design and the future pipeline of human skills.
Schaeffler prepares a large factory rollout
Humanoid’s agreement with Schaeffler covers an estimated 1,000 to 2,000 robots across global manufacturing sites by 2032. The first deployment is scheduled between December 2026 and June 2027 at two German locations.
Initial work includes handling boxes in Herzogenaurach and near-full-scale factory testing in Schweinfurt. Humanoid will help integrate the robots into existing production lines rather than requiring an entirely new facility.
The commercial relationship extends into components. Schaeffler is expected to become Humanoid’s preferred supplier for joint actuators through 2031, covering more than half the demand for its wheeled humanoid platforms and at least one million actuators.
That arrangement links deployment experience with the mechanical supply chain. Factory feedback can influence the components used in later robots, while a committed buyer gives Humanoid a path beyond small demonstrations.
Robots learn dexterity by observing workers
RLWRLD is collecting motion data in hotels, logistics facilities and retail settings. Cameras worn on workers’ heads and hands record tasks such as folding banquet napkins, lifting goods and arranging food displays.
Engineers convert footage into information about movement, joint angles and applied force. They add demonstrations captured with virtual-reality headsets, cameras and motion-tracking gloves, then use the combined data to train and teleoperate test robots.
Hand dexterity is a central challenge. Moving a cup, opening a box and placing an object on a conveyor require perception, grip control and continual correction. Human demonstrations contain practical details that are difficult to specify as rules.
RLWRLD expects industrial robots to reach scale around 2028. Hyundai plans to introduce Boston Dynamics humanoids at global factories beginning with Georgia in 2028, while Samsung aims to make all manufacturing sites AI-driven by 2030.
Worker participation must be part of deployment
Labour groups in South Korea have warned that robot adoption may affect employment and weaken the development of skilled workers. Motion data creates another concern: employees are not merely operating alongside machines but providing the examples used to automate tasks.
Organisations need clear consent, limits on how recordings are reused and rules governing retention and access. Workers should understand whether data evaluates them personally, trains a general system or supports one specific process.
Deployment plans should also distinguish between task automation and job replacement. At Lotte Hotel Seoul, current humanoids would take several hours to clean a room that a person completes in about 40 minutes. Back-of-house preparation may be more tractable than direct guest interaction.
That gap suggests near-term value will come from selected repetitive or physically demanding tasks. Human expertise remains necessary for exceptions, interpersonal service and adapting work when the environment changes.
Manufacturers should involve employees in task selection, safety testing and redesign. Their knowledge can identify subtle risks that engineers or managers miss, while transition plans can direct productivity gains toward safer and more skilled roles.
The road to factory-scale humanoids is being built through contracts, components and vast amounts of observed human practice. Technical progress will depend on how well robots learn those movements. Sustainable adoption will depend on whether the people who supplied that knowledge have a meaningful voice in how the resulting machines enter the workplace.
