ABB and Psyonic Team Up on Dexterous Robots

ABB Robotics has partnered with Psyonic, a California-based bionics company, to build industrial robots that can handle objects with something closer to human dexterity — and to speed up the rollout of 'physical AI' in manufacturing and logistics. The collaboration targets one of the hardest problems in the field: teaching a machine to grasp, grip and manipulate everyday objects the way a person does without thinking about it.

The clever part is where the training data comes from. Psyonic contributes its Ability Hand, a device originally developed as a prosthetic, which combines touch sensing, vibration feedback and articulated finger movement. People use the same hand to perform tasks, generating high-fidelity recordings of movement, contact and grip force — and that data is then used to train robotic systems to do the same jobs. 'Dexterous manipulation is ultimately a data challenge as much as a hardware challenge,' said Psyonic founder and CEO Aadeel Akhtar. 'By using the same Ability Hand on people and on robots, we can capture high-fidelity real-world data... then use that to train robotic systems more effectively.'

From pre-programmed arms to physical AI

Physical AI means systems that perceive, reason and act in the real world — robots that adapt to changing conditions instead of blindly following a fixed program. 'Human dexterity and the instinctive understanding of how to handle different objects is one of the most difficult things to replicate in industrial-grade robotics, but it's a fundamental need for truly autonomous and versatile robots,' said Marc Segura, president of ABB Robotics. The companies plan to integrate the technology with ABB's GoFa collaborative-robot platform and test it across automotive, aerospace, packaging, logistics and life sciences. It fits a broader industry trend: while generative AI was trained on internet-scale text and images, robotics needs real-world physical data, and that data is far harder to collect.

Why this matters for you

For the last few years, the visible face of AI has been text on a screen — chatbots, coding assistants, image generators. This partnership is a marker of the next phase, where AI moves its hands. The bottleneck isn't a smarter model; it's data about the messy physical world, and the workaround here — let humans wear the same hand the robot will use, then learn from how they move — is a genuinely neat way around it. My honest take is that physical AI is still earlier and slower than the chat-window version, and a single partnership won't change a factory floor tomorrow. But the things that get reliable robotic hands first — warehouses, packaging lines, assembly — are exactly the jobs people keep asking whether automation will reach, and progress on dexterity is how it eventually does.