AVITA launches Human-in-the-Loop physical AI, combining remote operation and autonomous control

prtimes.jp reported on September 15, 2026, that AVITA had launched Human-in-the-Loop physical AI, combining remote operation and autonomous control. We could not verify the information against primary sources, so please refer to the original report for details.

AVITA has announced an approach to bringing robots into workplaces that does not assume full autonomy, but instead includes remote operation. In workplaces such as factories, logistics operations and stores, where exceptions are common, the conditions for adoption shift from the maturity of autonomous capabilities to how much people can supplement them. This directly affects how workplaces prepare to adopt the technology.

What is new in this announcement is that AVITA has outlined a framework for extending the remote-operation capabilities it developed through avatar customer service directly to robots performing real-world tasks. Rather than the conventional approach of deploying robots only after achieving full autonomy, the framework assumes human involvement from the outset, with robots introduced into workplaces and their autonomy improved as data is collected. This is less an addition to the product lineup than a redefinition of the conditions for adoption. Read alongside our related articles, AVACOM ranks No. 1 in Japan for the number of deployed terminals and AVITA unveils AVITA OS, a software platform for safe collaboration between avatars and robots, this announcement appears to be a continuation of efforts to connect its remote customer-service operations and software platform to robots, rather than a sudden move. AVACOM's recognition points to accumulated experience in operations where people perform tasks remotely, while AVITA OS was positioned as a platform supporting human-machine collaboration. This physical AI initiative brings those two elements into real-world work and appears to mark a turning point in AVITA's business, expanding from “avatars on screens” to “physical agents performing work.” From an industry-structure perspective, the focus is not just on the robots' capabilities. Operational design that integrates exception handling at worksites, remote operation, work-log collection and the application of those logs to learning could become a basis for competition. Since the target sectors include factories, logistics operations and stores, what is needed is not a single high-performance robot but a system that can collect data and update controls for each type of work. Assuming that people will provide support can lower barriers to adoption, while the extent to which human labor can be reduced is likely to determine profitability. Unresolved questions include which tasks will be tackled first, the scale of the initial deployment in terms of robot numbers and locations, how robot models and control methods will be combined, and how often collected data will be incorporated into learning. It will also be necessary to determine how many people will be needed to operate robot fleets and how much the division of responsibilities with AVITA OS will be standardized.

In Japan, operational designs for remote operations and data collection at factories, logistics sites and stores are likely to be an important issue. Under the approach AVITA has outlined, systems that support operational platforms and the ongoing handling of learning data matter more than the robots themselves. This has implications for companies with expertise in related control software and workplace implementation.