Industrial Next and BMW Are Testing Robots That Switch Tasks in Minutes
Industrial Next announced an evaluation with the BMW Group Technology Office for AI-powered robotic handling in manufacturing. The company disclosed the agreement on the Startup Battlefield stage at TechCrunch Disrupt 2026. TechCrunch
The San Francisco company builds AI-driven, self-contained robotic stations. The stations run on proprietary foundation models, large AI systems trained on broad data that connect vision, task context, and motion. The pitch is direct. A production line could be retooled in minutes rather than weeks, without traditional programming.
Changeover time shapes what factories can automate. When reprogramming takes weeks, automation fits high-volume products that rarely change. When it takes minutes, lower-volume work and higher product mix become practical.
Industrial Next targets complex and delicate tasks still done by human labor. It says the system handles the loop of seeing, reasoning, and control inside the station itself, rather than as an add-on vision module or a fixed script of moves. The BMW work is framed as an evaluation, not a deployment or purchase.
Industrial Next was founded in 2021 by Lukas Pankau and Allen Pan. Both previously worked at Tesla. Pankau served as a lead electrical architect at Tesla before moving to Waymo. Pan led Tesla's autonomous factory efforts and later became a chief architect at autonomous driving company Neolix. TechCrunch
Separately, the company operates a 10,000-square-foot Real2Sim2Real Arena for Physical AI in San Francisco, described in a Sept. 29 post. Industrial Next on LinkedIn Its official domain is industrialnext.ai.
The broader context here will be familiar to manufacturing engineers. Dexterous manipulation, careful handling of varied objects, remains the hard boundary for factory robotics. Picking known rigid parts from fixed fixtures is solved. Handling deformable materials, tight tolerances, varied lighting and part presentation, and doing it at takt time, the pace needed to keep a line running, is not. Foundation models that combine vision with task context allow natural task description instead of explicit motion code, but they raise questions around determinism, validation and failure modes that factories cannot waive.
In my view, the station design is as notable as the BMW name. Self-contained stations lower integration risk. They keep sensing, inference and control local to the cell, which helps with response time and with isolating a new process from the rest of the line. The tradeoff is familiar. Cells that are easy to drop in still have to connect to upstream and downstream equipment, material flow, safety systems and quality logging. Autonomy inside the station does not remove systems work around it.
Looking at what this means for plant teams, the evaluation structure makes sense. It lets both sides test grasp success, recovery behavior, changeover workflow and operator burden on real parts without committing to production targets. Worth flagging, the facts available do not specify throughput, accuracy, safety certification or how the model is updated when parts or lighting change. Those details will determine whether minutes-long retooling holds up outside a staged demonstration.
There is also a workforce point to state plainly. Automating delicate manual tasks changes the content of line work. It shifts human effort from direct handling toward supervision, exception handling and process definition. I have watched my own children adopt new tools faster than the institutions around them, and factories often follow the same pattern. Operators adapt quickly when the tool is legible. The open question is whether programming in natural task terms stays legible and debuggable on the floor.
If the approach transfers to production, the payoff is flexibility. Shorter changeovers allow smaller batches, faster product iterations and reshoring of work that was too fiddly to automate economically. That long arc is worth watching, even while this specific partnership remains at the evaluation stage.
