PSYONIC's FDA approved Ability Hand has been capturing real world touch and motion data for years. ABB Robotics is now using it to teach industrial robots to handle objects current grippers cannot.
For years, the people who use PSYONIC's Ability Hand have been doing something most of them never signed up for: generating a steady stream of real-world touch and motion data that industrial robotics has struggled to replicate on its own. That data is now flowing into ABB's factory robots, in a partnership that exposes a feedback loop between assistive technology and industrial automation that is older and more interesting than the press release makes it sound.
PSYONIC, a U.S. startup founded in 2015, makes a multi-articulating prosthetic hand with pressure sensors, vibration feedback, and compliant fingers. The device is FDA-approved, covered by Medicare, and used by more than 300 patients, with early purchasers including Meta. ABB Robotics, the industrial robot arm of ABB Group, is now working with PSYONIC to take the dexterity lessons embedded in that population and apply them to one of the most stubborn problems in factory automation: teaching a robot to pick up, manipulate, and place objects that are not pre-positioned and not identical. The two companies are pairing ABB's GoFa collaborative robot arm with PSYONIC's Ability Hand to collect and use touch and motion data from prosthetic users, according to The Robot Report.
The underlying bet is that prosthetic users are, in effect, a daily dexterity data-collection program. Every time someone with an Ability Hand picks up a cup, ties a shoe, or handles a tool, the hand's pressure and motion sensors capture a small, physical-AI-ready example of how a human actually solves a manipulation problem. Physical AI is the category name for AI that learns from and acts in the physical world through sensors and actuators, as opposed to text or images. ABB frames this work as part of its push toward Autonomous Versatile Robotics, a category label for robots that can sense, reason, move, and precisely handle objects in dynamic settings. Marc Segura, president of ABB Robotics, told The Robot Report that 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, and a fundamental need for truly autonomous and versatile robots. That language is ABB's own marketing frame, and is best read as a category the company has chosen to compete in, not as a settled technical claim.
The constructive read is that the prosthetic field has been quietly solving manipulation problems for years in a hard, real-world loop, and that knowledge is now flowing back into general robotics. The less constructive read is that the gap between works in a prosthetic user's daily life and works on a factory floor is real, and not yet closed by an announcement. Existing factory grippers, mostly suction cups and parallel-jaw fingers, have well-known limits: they depend on tool changes, they struggle with soft, deformable, or irregularly shaped objects, and they assume the world around them is rigid. Dexterous manipulation has been the part of the problem that has not really been solved at scale.
There is also a timing tell that the press release does not foreground. PSYONIC's own CEO has told The Robot Report that the company's industrial-side business has only recently flipped to outweigh the prosthetic side, and that the shift tracks the build of physical-AI hype over the past year. That is a useful tell for a reader weighing the announcement: the partnership is real, but the commercial timing is also riding a wave.
The SoftBank context belongs in the background, not the lede. ABB Group sold its Robotics division to SoftBank for $5.3 billion in October 2025, which helps explain why the new partnership has a sharper edge. An industrial robotics business that is now a standalone, capital-backed unit has a stronger reason to chase a frontier category like physical AI than a division inside a slower-moving conglomerate might have had a year earlier.
What to watch: whether the data ABB collects from Ability Hand users translates into measurable gains on factory tasks current grippers cannot handle, whether PSYONIC's industrial business outpaces its prosthetic business on revenue, and whether independent customer-side testing confirms that prosthetic-trained manipulation generalizes off the prosthetic context. The press release announces a feedback loop. The interesting question is whether the loop closes.