Industrial Next says its AI robot stations retool lines in minutes, and announces BMW Group Technology Office partnership to evaluate its dexterous manipulation

Lukas Pankau, Industrial Next's co-founder, put the problem in one sentence. "But I couldn't figure out how to program a robot, and this was pretty common for engineers at Tesla," Pankau said ^[6].

Here's what plants actually do, by Pankau's account. Traditional industrial robots are programmed for highly specific, repeatable tasks, so a small change in the process or product can make them fail or trigger costly downtime ^[7]. Manufacturers respond by simplifying their products for repeatability ^[7]. The pitch for a flexible robot is that designers can stop doing that, and a pilot is the only place that gets tested.

The obstacle is data. AI models can be trained on the internet, but robotics has no comparably large body of data for teaching machines complex physical tasks ^[8]. Industrial Next's answer is simulation. Pankau said the company builds the robots' understanding of a process "using an extremely high-fidelity simulation engine that is close to the real dynamics to where we can let the robot do these kinds of high precision tasks almost without seeing the parts at all." ^[9] The minutes figure rests on that fidelity, and the company is its only source in the report ^[11].

TechCrunch's report does not give a timeline or a result for the BMW evaluation ^[1]. Industrial Next was founded in 2021, so the announcement comes about five years in ^[10].

For a plant weighing a call from the vendor, we'd sort lines on two axes: how often the product mix changes, and what a changeover costs in downtime today. Lines that change often and lose a lot of time per change are where a minutes-long retool would pay, and those are the only lines worth a pilot. Lines that rarely change have little to gain, and so do lines whose changeovers are already cheap.

For the first group, the forcing function is a timed changeover on a part the vendor has not seen, run on the buyer's floor and compared with the changeover time the plant records today. The tradeoff is that this puts the vendor on the hardest case first and may end the pilot early. We think an early end costs less than finding out after installation.