Related: Swiss startup Flexion shows a humanoid robot doing office tasks on its own.
BMW has moved its humanoid robot experiment from a demonstration into the everyday grind of factory logistics. At the automaker’s plant in Spartanburg, South Carolina, Figure AI‘s newest robot, Figure 03, has started handling a real material handling job, taking parts that arrive jumbled in large containers and sorting them into the exact order the assembly line needs. It is an unglamorous task, and that is the point, because the boring, repetitive work of moving and sequencing parts is where a general purpose robot has to prove it can actually earn its keep.
The specific job is what the industry calls sequencing. Components show up unsorted in bulk bins, and Figure 03 picks them out and arranges them into a sequencing trolley in the order they will be installed. That trolley then goes to a collection point, where an automated tugger train or a smart transport robot carries the parts to the line so they arrive just in sequence for the workers fitting them. It is the kind of dexterous, judgment laden task, see the part, identify it, grasp it, place it correctly, that is easy for a person and has been stubbornly hard for machines, which is exactly why it is a meaningful test of a humanoid.
Figure 03 is not BMW’s first robot at the plant. The partnership began with an earlier model, Figure 02, which BMW says supported the production of more than 30,000 X3 vehicles over about ten to eleven months, working in the body shop to insert sheet metal parts for welding, a fast and physically demanding job. Moving from that single repetitive insertion task to the more varied work of logistics sequencing is a deliberate escalation, a sign that both companies think the robot is ready for jobs that require more flexibility and less scripting.
The newer robot was built with a factory floor in mind. Brett Adcock, Figure’s chief executive, has highlighted a set of practical upgrades on Figure 03, including soft exterior components meant to make it safer to work near people, wireless charging so it can top up without downtime, and speech to speech communication for more natural back and forth with human coworkers. Its redesigned hands carry tactile sensors and cameras built into the palms, which give it the finer touch and better object recognition that picking and placing loose parts demands. Those are the unshowy engineering details that separate a stage demo from a machine that can run a shift.
For BMW, the deployment fits a broader push it markets as its iFACTORY approach, folding artificial intelligence and virtualization into production, with robot development running in parallel at Spartanburg and at Figure AI. For Figure, putting its robot into a marquee customer’s real operations is the validation the whole humanoid industry is chasing, proof that these machines can do useful, sustained work outside a lab rather than perform impressive one off tricks. A logistics task inside a working car plant is about as real as that test gets.
The honest caveat is that a deployment is not the same as widespread replacement, and one robot sequencing parts at one plant is an early step rather than a finished transformation. Humanoids still cost a lot, and factories are unforgiving environments where reliability across thousands of hours, not a good day, is what matters. The questions worth watching are how consistently Figure 03 performs shift after shift, how much human oversight it still needs, and how quickly the economics make sense at scale. What is clear is the direction, with a growing number of carmakers and logistics firms testing humanoids on exactly this kind of work, and BMW moving its own program from a single task to a more demanding one. The robots are no longer just auditioning, they are starting to clock in.
Related on Entrelligence: Flexion’s autonomous humanoid, and Amazon’s next-generation Zoox robotaxi.