Industrial Next wants factory robots to switch tasks in minutes, not weeks

Industrial Next, a San Francisco-based startup, presented its pitch for making industrial automation more flexible on the Startup Battlefield stage at TechCrunch Disrupt 2026. According to the article, automation in most factories is fixed: robots and machines are programmed for a specific task and product, and changing a line requires physical reprogramming and retooling, or replacing equipment, which can take weeks or months.

The company builds autonomous, AI-driven robotic stations based on its own foundation models that connect vision, task context and motion. It says this allows a production line to be reconfigured in minutes rather than weeks, without traditional programming. The goal is to let manufacturers automate complex, delicate tasks that until now have required human labor. On stage, the company announced a collaboration with the BMW Group Technology Office to evaluate its AI-based dexterous manipulation in manufacturing processes.

Industrial Next was founded in 2021 by Lukas Pankau and Allen Pan, both Tesla alumni. Pankau was lead electrical architect at Tesla before moving to Waymo; Pan led Tesla's autonomous factory efforts and later served as chief architect at autonomous driving company Neolix. The article says the idea grew out of their frustration at Tesla: the cars and self-driving featured advanced technology, but the factory relied on equipment that Pankau describes as being from the 1980s, hard to program even for very capable engineers.

Pankau explains that traditional industrial robots are programmed for very specific, repeatable tasks, so a small change in the process or product can cause failures or costly stoppages, which forces products to be simplified. Making robots truly flexible requires huge amounts of training data, and robotics lacks the large datasets that exist for models trained on the internet. According to Pankau, it is possible to make robots dance because there is data for that, but not to have them assemble high-precision parts.

Their answer is simulation: the robots practice in a very high-fidelity virtual environment that closely reproduces real-world physical dynamics, which, according to Pankau, allows them to perform high-precision, high-dexterity tasks having barely seen the real parts. The stations can be reconfigured, combined and reused across product lines, and new tasks are taught through simulation and demonstration rather than hand-coded scripts.

Pankau says some of its stations are already being used in the manufacturing of one of the world's most popular smartphones (he could not disclose the customer). They handle tasks that manufacturers had spent years trying to automate without success, such as threading a delicate flex cable or snapping on the cover of a tiny connector. He also described a "film peeling" station, where robots remove the protective film from camera modules and buttons at a speed equal to or faster than humans and with 100% accuracy, even when transferred to a phone model and production line they had not seen before. These figures are the company's own claims and are not independently verified in the article.

The stated ambition is to move beyond phones toward more complex tasks and, ultimately, to build an entire car from start to finish without any human touching it.