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

In astir factories, automation is fixed: robots and machines are programmed for 1 circumstantial task and product. Changing a enactment means reprogramming and carnal retooling, oregon swapping retired equipment, which tin instrumentality weeks oregon months. San Francisco-based Industrial Next says it’s changing that.

The startup, which presented contiguous connected the Startup Battlefield signifier astatine TechCrunch Disrupt 2026, builds AI-driven, self-contained robotic stations powered by proprietary instauration models that link vision, task discourse and motion. The institution says this lets a accumulation enactment beryllium retooled successful minutes alternatively than weeks without accepted programming. The extremity is to fto manufacturers automate analyzable and delicate tasks that traditionally necessitate quality labor.

The institution announced connected signifier that it’s partnering with BMW Group Technology Office to measure Industrial Next’s AI-powered dexterous manipulation for manufacturing.

Industrial Next was founded successful 2021 by Lukas Pankau and Allen Pan, some of whom antecedently worked astatine Tesla. Pankau served arsenic a pb electrical designer astatine Tesla earlier moving connected to Waymo. Pan led Tesla’s autonomous mill efforts and aboriginal became a main designer astatine autonomous driving institution Neolix.

The brace came up with the thought for Industrial Next portion moving unneurotic astatine Tesla. Pankau’s archetypal brushwood with Pan was virtually tripping implicit him arsenic Pan was dormant connected a mill catwalk aft pulling an all-nighter. Although the 2 didn’t get on astatine first, they soon realized they shared an obsession with getting things done and had complementary skills. 

After years of watching cutting-edge conveyance and self-driving exertion get held backmost by decades-old mill equipment, the founders realized that manufacturing needed to beryllium rebuilt.

Image Credits:Industrial Next /

“We had each these fancy things that we could bash connected cars. I could tally instrumentality learning algorithms to plan a car for me. But I couldn’t fig retired however to programme a robot, and this was beauteous communal for engineers astatine Tesla,” Pankau said. “You had this monolithic divide, right? You had radical who are truly bully astatine what they’re doing–designing cars and designing self-driving systems–and past you had them efficaciously stumped astatine the mill looking astatine this instrumentality from the 1980s. And truthful we started rethinking what this should look similar utilizing the exertion that we person contiguous and utilizing everything that we cognize today.”

The 2 kept reasoning astir their frustrations with concern instrumentality adjacent aft moving connected to their adjacent jobs. They stayed successful touch, often reminiscing astir the machines that ne'er functioned the mode they wanted, and yet decided they could bash better.

“I was astatine Waymo astatine the time, benignant of bored retired of my caput due to the fact that I was missing that gait of Tesla,” Pankau said. Pan, helium added, was successful a akin presumption astatine Momenta, a developer of self-driving systems successful China. Pan recalled their conclusion: “We’re like, we’re not getting immoderate younger, and we’re inactive evidently passionate astir it. We inactive support complaining astir concern technology. Like, this is 3 years later, we should commencement the company.”

Pankau notes that accepted concern robots are typically programmed to execute highly specific, repeatable tasks. As a result, adjacent a tiny alteration successful the process oregon merchandise tin origin the robot to neglect oregon trigger costly downtime, which forces manufacturers to simplify their products for repeatability.

Making robots genuinely flexible presents a antithetic challenge. They request tremendous amounts of grooming information and, dissimilar AI models trained connected the internet, robotics lacks the immense datasets needed to thatch machines however to execute analyzable carnal tasks. 

Image Credits:Industrial Next /

“We tin marque robots bash each these chill things similar creation around, due to the fact that there’s information for that, but we can’t marque them assemble high-precision things,” Pankau explained. “This is wherever we’ve made a batch of advancement connected utilizing simulated environments to bash these high-precision and high-dexterity tasks. The concealed condiment for the robots that we person is that we person fixed them an knowing of the process successful general, but we physique that knowing utilizing an highly high-fidelity simulation motor that is adjacent to the existent dynamics to wherever we tin fto the robot bash these kinds of precocious precision tasks astir without seeing the parts astatine all.”

In different words, the robots signifier successful a highly elaborate virtual satellite that intimately mimics real-world physics. Industrial Next’s robotic stations tin beryllium reconfigured, combined, and redeployed crossed merchandise lines. New tasks are taught done simulation and demonstration, alternatively than hard-coded scripts. 

Pankau explained that immoderate of the startup’s workstations are already utilized successful making 1 of the world’s astir fashionable smartphones. They grip tasks that manufacturers had tried and failed to automate for years, specified arsenic threading a delicate flex cablegram and snapping a tiny connector headdress into place. (He wasn’t astatine liberty to disclose the customer.) He besides described a “film peeling” station, wherever their AI-powered robots portion protective movie disconnected camera modules and buttons astatine oregon supra quality velocity with 100% accuracy, adjacent erstwhile moved to a caller telephone exemplary and accumulation enactment they person ne'er seen.

“We started this institution due to the fact that that is the benignant of show that we privation we had astatine Tesla, wherever we were bypassing robot stations due to the fact that we couldn’t programme them efficaciously capable utilizing accepted technologies,” Pankau said. “Now we privation to instrumentality this to much analyzable tasks than conscionable gathering a phone. We yet privation to instrumentality this to gathering a afloat car end-to-end without immoderate quality touching ever.”

To cheque retired much of the Startup Battlefield — on with a slate of established founders and VCs whose takes connected the manufacture you won’t privation to miss — join america astatine Disrupt, happening this week successful downtown San Francisco.