For decades, factory automation has suffered from a blind spot. Industrial robots are exceptionally good at doing the exact same thing millions of times in a row, provided nothing changes. Shift a part by two millimeters, and the whole assembly line grinds to a halt.
Fixing that brittleness is what drove Rudy Cohen, Albane Dersy, and Louis Dumas to launch Inbolt in Paris in 2019. Instead of forcing factories to re-engineer entire production lines to accommodate rigid machines, the French startup decided to give robotic arms real-time sight and spatial awareness. Meanwhile, you can read similar stories here: The Ghost in the Silicon and the Bet That Defies Giants.
The Problem With Blind Automation
Traditional industrial robotics relies on hardcoded trajectories. If you buy a robotic arm from manufacturers like FANUC, KUKA, or ABB, it executes a predetermined path written in code. It doesn't know what it's touching. It just assumes the workpiece is waiting at the exact coordinate specified in the program.
Real-world manufacturing is messy. Parts warp, bins arrive with jumbled components, and assembly tolerances drift. For years, solving this meant installing massive, expensive overhead 3D camera rigs and spending weeks on calibration. To see the full picture, we recommend the recent analysis by Gizmodo.
Factory engineers often spend more time troubleshooting rigid automation cells than the systems save. When a line stops because a part is slightly out of place, downtime costs thousands of dollars a minute. Traditional machine vision systems were too slow to correct these errors on the fly, operating with lag times that made real-time trajectory adjustments impossible.
How Inbolt Changed the Equation
Inbolt sidestepped these limitations by rethinking where the vision system lives. Rather than mounting bulky cameras on factory ceilings, they attach a standard 3D camera directly onto the robot's wrist, turning the arm into an active seeker.
The core software, known as GuideNOW, pairs this on-arm camera with proprietary artificial intelligence trained directly on CAD models. When the robot approaches a workstation, the system processes 3D data in under 80 milliseconds. It calculates the exact position and orientation of the target component, corrects the robotic path on the fly, and guides the tool home.
You don't need massive datasets or days of machine learning training. You upload the CAD file of the part, and the software is ready to deploy. This approach cuts commissioning times from weeks down to minutes.
Real Production Floors Tell the Story
Theory is nice, but factory managers care about uptime and scrap rates. Inbolt's deployment inside major automotive plants like Stellantis demonstrates the financial impact. By retrofitting existing robotic stations with real-time vision guidance, Stellantis eliminated complex mechanical re-engineering and saved millions on high-maintenance welding and assembly cells.
In unstructured bin-picking applications—traditionally one of the hardest challenges in industrial automation—the system achieves high success rates at speeds under one second per pick. The robot handles tangled heaps of components, figures out how to grasp them, and adapts if a part slips mid-air.
This hardware-agnostic capability means factories don't have to rip out existing investments. They can bolt the camera onto legacy machinery, load the software, and upgrade a dumb machine into a vision-guided worker.
Scaling Up and Moving Forward
The market has noticed. Backed by institutional rounds totaling over €20 million—including investments from Exor Ventures, the backer of Ferrari and Stellantis—Inbolt is expanding rapidly across Europe, the United States, and Japan.
The company's integration into hardware kits like the NVIDIA-powered Universal Robots AI Accelerator signals a shift toward making vision-guided robotics the default standard rather than a custom engineering project.
If you're running a manufacturing line today, stop looking at automation as an all-or-nothing proposition. You don't need a brand-new factory floor to make your operations flexible. Retrofitting legacy arms with real-time 3D vision bridges the gap between rigid hardware and the chaotic reality of production. Audit your most troublesome assembly cells, look for bottlenecks caused by part variation, and test software-driven vision layers that can integrate with the equipment you already own.
