Pilot projects in hospitals and early efforts in construction are pushing the limits of “physical AI”
Humanoid robots are moving from science fiction to pilot projects in hospitals, warehouses and, potentially, construction sites, raising new questions for health and safety professionals about how “physical AI” could change the risk landscape at work.
Modar Alaoui, general partner at ALM Ventures and chair of the Humanoids Summit in Silicon Valley, describes humanoids as a “physical manifestation of AI” that is starting to move from the cloud into the three-dimensional environments where people work. He argues that while the technology is still in its early stages, its safety potential is significant.
From “physical AI” concept to workplace pilot projects
Alaoui’s core thesis is that most workplaces are already designed for human bodies: door handles, stairs and counter heights all assume a roughly human form factor. That makes humanoid robots a natural fit to operate tools, navigate buildings and perform tasks in existing facilities without redesigning infrastructure. In his view, generative AI and modern robotics together give “almost a new soul to robotics,” making it possible to automate more complex physical work.
The safety opportunity, he says, sits at the base of the hierarchy of needs. “Safety is always going to be at the base of everything,” Alaoui notes, adding that he believes physical AI “can be trained to add more safety to human lives, to the workplace, to manufacturing.” For health and safety leaders, that could mean a new class of tools to remove workers from high-risk, repetitive or ergonomically damaging tasks.
Hospitals offer an early proving ground for safety gains
Healthcare has emerged as one of the earliest testbeds. Alaoui points to Emancro, a portfolio company of his venture fund that is building a highly specialized humanoid for hospital pharmacies. In a geo‑fenced pharmacy area, the robot is designed to handle medications: taking drugs from storage, scanning them, placing them correctly, opening drawers and tracking how long a medication has been out, with automatic disposal if it exceeds a time limit. The goal is to reduce human error and shrink “loss” in a tightly controlled, highly regulated micro‑environment. Early pilots are underway with major U.S. health systems including Mayo Clinic and Johns Hopkins.
Another early application involves hospital logistics. Companies such as Diligent Robotics are deploying humanoids that function like a “robot butler,” delivering medications and supplies to specific rooms, at set times, and potentially taking vital signs in the future. Alaoui suggests that, at least initially, public acceptance may be higher for robots that transport medications than those that administer them directly, given the liability and trust issues involved.
These pilots foreshadow similar patterns in heavy industry: strict entry barriers, but large potential once technologies clear regulatory and cultural hurdles. Alaoui notes that in the United States, OSHA only tends to come into focus for some builders after accidents occur, drawing a parallel to how autonomous driving technology outpaced regulatory and social readiness in its early years. He expects a similar lag between technological possibility and formal safety frameworks for humanoids.
Construction and heavy industry eye high‑risk tasks for robots
For sectors such as construction, mining, and energy, he sees a clear role for humanoids in dangerous, labour‑intensive work. “Everything that we know today in construction is labour intensive. Human labour intensive. There’s not much automation at all,” Alaoui says. Humanoid robots, he argues, could carry materials, operate on upper floors and work in confined or unstable areas where falls, crush injuries and exposure to hazards are common.
Several start-ups are already targeting these scenarios, including companies developing humanoids to take on electrical tasks such as pulling wire through conduits and activating electrical systems. In some cases, these machines may act as “co‑electricians” working alongside tradespeople, with the potential to shift workers toward more supervisory or diagnostic roles instead of hands‑on exposure to live systems.
Still, Alaoui stresses that the field is at an early stage. “It’s all promising,” he says, but questions remain about how long widespread deployment will take, under what conditions, and what the bottlenecks will be. Adoption will depend not only on technical performance and cost, but also on regulatory clarity, labour availability, union positions and public comfort with robots working near people.
For health and safety professionals, the message is less about immediate replacement of workers and more about preparation. As humanoid pilots expand from hospitals and warehouses into construction and heavy industry, safety leaders may soon be asked to evaluate new kinds of risk assessments, procedures and training—this time not just for people working with machines, but for people working alongside them.
