Humanoids, Embodied AI and the Blurring of Job Roles in Systems
Written by Dr Alan Burden, QUT, Postdoctoral Research Fellow in Designing Socio-Technical Systems program
The next stage of collaborative robotics is likely to arrive in the form of humanoids, with Artificial Intelligence (AI) speeding up their development. For those people sharing a factory floor or building site with these machines, that shift will change their work and reach into the teams, workplaces and industries around them. Generative AI has been reshaping what we’ve colloquially known as “white-collar” work. Humanoid robots have recently been considered its physical, embodied AI counterpart, and are likely to reshape blue-collar work too.
Much of the public attention still sits with humanoids walking, running, and dancing. Less visible is how quickly the long-standing problems underneath are being solved. The technology has three layers: cognition, the language model that understands an instruction; skills, the learned behaviours; and coordination, the body control that makes the action happen (Hurst, 2026). This ‘stack’ is a system that is familiar to software developers and is the framing used by Jonathan Hurst, co-founder and Chief Robot Officer of the humanoid company Agility Robotics. Cognition is “the easy AI”, because its training data already exists online; the body is harder, and it is catching up (Hurst, 2026). Humanoid robots are where these layers meet, and two overlapping terms have emerged to describe it: embodied AI, which emphasises intelligence that learns through a body; and physical AI, an industry-led term for AI that acts in the physical world.
What humanoids will actually do remains open. One global industry body, the International Federation of Robotics put it plainly that “if and when a mass adoption of humanoids will take place remains uncertain” (International Federation of Robotics, 2025). Humanoids are more likely to fill the activity gaps between specialty robots, such as the robotic welding cell, the autonomous mobile robot moving pallets and the inspection drone. These machines will keep doing what they were designed for, with better ways for people to direct and communicate with them. Humanoids are expected to “complement and expand upon existing technology” rather than replace it (International Federation of Robotics, 2025). Some tasks will overlap with other robots, and their value lies less in any single task than in moving between many general-purpose ones (Hurst, 2026).
Early deployments seem to support this use of humanoids. Car manufacturer BMW’s Spartanburg plant in South Carolina uses humanoids from the start-up company Figure AI. The first Figure robots began by loading sheet metal; its newer F.03 robot moves between parts sequencing, sorting and pulling carts down the line (Figure AI, 2025, 2026). At a warehouse run by logistics company GXO, Agility Robotics’ Digit has moved more than 100,000 totes, the plastic crates used to move stock, working between conveyors, mobile robots and floor stacks (Agility Robotics, 2025). Neither robot is tied to a single job. Car manufacturer Hyundai’s plan for Atlas, built by its robotics company Boston Dynamics, assigns robots labour-intensive and high-risk tasks, “allowing human workers to focus on training the robots and providing oversight” (Hyundai Motor Group, 2026).
Instructing these machines is getting easier. As the way we interact with computers has changed, so has the way we interact with robots. For much of the public, chatbots built on large language models (LLMs) were a first real encounter with AI, and they shifted expectations of how we talk to our devices. AI now sits inside everyday tools, responds to voice and acts through other agents. Dictation is having a resurgence, as AI can interpret spoken instructions in context far better than earlier transcription tools ever could.
Robots, too, are increasingly taught rather than programmed line by line. The latest model from Physical Intelligence, a robotics AI start-up, can “compose and recombine the skills it learned to solve new tasks”, including learning an unfamiliar kitchen appliance through spoken coaching (Physical Intelligence, 2026). Physical Intelligence is not alone. NVIDIA, best known for the chips that power AI, now builds robot models of its own and claims its GR00T N2 handles new tasks in new environments more than twice as often as rival models (NVIDIA, 2026). With chipmakers, start-ups and car manufacturers all chasing the same capability, robot intelligence is fast becoming a crowded and competitive field.
Perhaps more importantly, what one robot learns no longer stays with that robot. In one demonstration, a robot folded laundry without ever being shown the task, drawing on data collected with a different robot (Physical Intelligence, 2026). The individual stack becomes part of a network of machines.
Plain language also lowers the barrier on site. In a simulation study, researchers at the University of Michigan directed a robot to install plasterboard through everyday instructions, with more than 99% of instructions interpreted correctly (Park et al., 2024). Once an operator, fitter or inspector can direct a robot, long-standing roles begin to converge.
Software development felt this change first. By 2026, 73% of coding assistant and agent users used them every day (Stack Overflow, 2026). As Elizabeth Stone, Chief Product and Technology Officer at Netflix, puts it, “the days of very narrow, deep specialization feel more limited” (Rachitsky, 2026).
Changes of this scale are not new. Railways and motor vehicles displaced whole trades while creating roles that had no name beforehand (Bogart, 2014; Manyika et al., 2017). What differs now is reach: few professions are likely to be untouched. That rarely means replacement, but more often, the role is redrawn or merged into other linked roles.
Thinking in systems is one skill likely to rise in value. The World Economic Forum expects 39% of workers’ core skills to change by 2030, naming robots and autonomous systems as a driver of demand for “programming and systems thinking” (World Economic Forum, 2025). Even the factory is being reframed as “a robot that’s orchestrating robots to build things that are robotic”, in the words of NVIDIA chief executive Jensen Huang (NVIDIA, 2025).
Our own work at the Australian Cobotics Centre points the same way. When colleagues and I mapped dependencies across the robot, product, operator and workplace, two of the three most connected factors belonged to the workplace: its scope of application and its degree of automation (Guertler et al., 2023).
Human judgement, care and responsibility for safety remain essential, now shared across capable teams, many of whom will never have “engineer” in their title. Robots strong enough to do useful work are strong enough to cause injury (Hurst, 2026), so safety belongs in the design. Getting the robot working may turn out to be the easy part. The harder, more interesting work is understanding the whole system around it.
References
Agility Robotics. (2025, November 20). Digit moves over 100,000 totes in commercial deployment. https://www.agilityrobotics.com/content/digit-moves-over-100k-totes
Bogart, D. (2014). The transport revolution in industrialising Britain. In R. Floud, J. Humphries, & P. Johnson (Eds.), The Cambridge economic history of modern Britain: Vol. 1. 1700–1870 (pp. 368–391). Cambridge University Press. https://doi.org/10.1017/CHO9781139815017.014
Figure AI. (2025, November 19). F.02 contributed to the production of 30,000 cars at BMW. https://www.figure.ai/news/production-at-bmw
Figure AI. (2026, June 30). F.03 arrives at BMW. https://www.figure.ai/news/f-03-at-bmw
Guertler, M. R., Brackemann, T., Burden, A., & Caldwell, G. (2023). Mapping socio-technical dependencies to enable the successful adoption of collaborative robots in industry. Procedia CIRP, 119, 564–569. https://doi.org/10.1016/j.procir.2023.02.150
Hurst, J. (2026, May 29). The future of humanoid robotics [Video]. TEDxPortland. YouTube. https://www.youtube.com/watch?v=21BzAy5YEuE
Hyundai Motor Group. (2026, January 5). Hyundai Motor Group announces AI robotics strategy to lead human-centered robotics era at CES 2026. https://www.hyundaimotorgroup.com/en/news/hyundai-motor-group-announces-ai-robotics-strategy–to-lead-human-centered-robotics-era-at-ces-2026
International Federation of Robotics. (2025, August 14). Humanoid robots: “Vision and reality” paper published by IFR [Press release]. https://ifr.org/ifr-press-releases/news/humanoid-robots-vision-and-reality-paper-published-by-ifr
Manyika, J., Lund, S., Chui, M., Bughin, J., Woetzel, J., Batra, P., Ko, R., & Sanghvi, S. (2017). Jobs lost, jobs gained: Workforce transitions in a time of automation. McKinsey Global Institute. https://www.mckinsey.com/featured-insights/future-of-work/jobs-lost-jobs-gained-what-the-future-of-work-will-mean-for-jobs-skills-and-wages
NVIDIA. (2025, October 28). NVIDIA GTC Washington, DC: Live updates on what’s next in AI. NVIDIA Blog. https://blogs.nvidia.com/blog/nvidia-gtc-washington-dc-2025-news/
NVIDIA. (2026, March 16). NVIDIA and global robotics leaders take physical AI to the real world [Press release]. https://nvidianews.nvidia.com/news/nvidia-and-global-robotics-leaders-take-physical-ai-to-the-real-world
Park, S., Wang, X., Menassa, C. C., Kamat, V. R., & Chai, J. Y. (2024). Natural language instructions for intuitive human interaction with robotic assistants in field construction work. Automation in Construction, 161, 105345. https://doi.org/10.1016/j.autcon.2024.105345
Physical Intelligence. (2026, April 16). π0.7: A steerable model with emergent capabilities. https://www.pi.website/blog/pi07
Rachitsky, L. (Host). (2026, July 19). Why Netflix is betting on systems thinkers, not specialists, in the AI era | Elizabeth Stone (CPTO) [Audio podcast episode]. In Lenny’s Podcast. https://www.lennysnewsletter.com/p/netflix-cpto-on-ai-and-the-future
Stack Overflow. (2026). 2026 Developer Survey: AI. https://survey.stackoverflow.co/2026/ai
World Economic Forum. (2025). The future of jobs report 2025. https://reports.weforum.org/docs/WEF_Future_of_Jobs_Report_2025.pdf