Hospitals are shifting focus from purchasing physical AI robots to building the governance, safety frameworks, and operational strategies needed to scale these systems safely. Leading experts in medical technology warn that treating intelligent machines as clinical infrastructure requires strict oversight before they can move from promising pilots to trusted daily use.
Extending the care team
None of the featured executives envision autonomous machines taking over clinical decision-making. Instead, they describe physical AI as a direct extension of the existing care team that handles routine physical tasks while clinicians retain full control. Davit Martirosyan, cofounder and head of engineering at Expper Technologies, said healthcare leaders should prioritize adaptive interaction systems for patients who struggle with standard screens, including children and older adults. He added that a physical system can maintain attention and adapt communication in ways that screen-based interfaces cannot match. The longer-term vision involves an interconnected ecosystem where specialized devices share context and safety rules instead of operating as isolated tools.
Coordinating digital and physical systems
David Niewolny, senior director of business development for healthcare and medtech at NVIDIA, expects hospitals to deploy coordinated teams of digital and physical AI agents that continuously exchange information. Digital agents will reason over patient data and orchestrate documentation, while physical systems handle logistics, monitoring, and procedural support. Dr. Filippo Filicori, system chief of surgical innovation at Northwell Health, sees intelligence becoming embedded directly inside surgical workflows. Future systems will combine video feeds, robotic kinematics, force feedback, and device telemetry to recognize surgical phases and detect developing complications. Filicori emphasized that those capabilities represent assistance rather than replacement. "The most consequential systems will probably not be the most visually impressive humanoid robots - they will be the systems that reliably remove delays, reduce variation and return time to clinicians," he said.
Building governance and safety layers
Because physical AI directly interacts with people and equipment, governance and safety oversight become foundational requirements. Martirosyan advised organizations to assign every device a secure identity, limit its permissions, and track exactly what actions it takes and when humans intervene. Deployments require multiple protection layers, including immediate human takeover, emergency stopping mechanisms, and predictable fallback modes during network failures. Filicori argued that responsibility for oversight must extend beyond IT departments to include clinical, operational, and administrative leadership. "CIOs should begin by treating physical AI as clinical infrastructure, not as another software application," he said. "A generative AI failure may produce an incorrect answer. A physical AI failure may move the wrong object, collide with a person, interrupt a procedure or injure tissue."
Validating through simulation and workflow design
Niewolny advocates for a simulation-first validation strategy that tests systems in virtual environments reproducing hospital conditions and potential failure scenarios before any patient contact. Simulation accelerates development but does not replace clinical evidence, hospital validation, or regulatory review. Leaders should also start by identifying operational bottlenecks rather than selecting hardware. "CIOs should start with the capacity and access problem, not with the robot," Niewolny said. "Where are clinicians stretched? Where are patients waiting? Where is expertise scarce?" Only after mapping those constraints should organizations determine whether physical AI fits the workflow. For teams managing AI for Healthcare initiatives, success depends less on acquiring advanced hardware and more on redesigning processes, establishing clear accountability, and preserving human judgment at every stage.
Why this matters for healthcare professionals
Hospital administrators and clinical staff will see physical AI integrated quietly into daily operations rather than arriving as standalone autonomous caregivers. Leaders who establish multidisciplinary governance boards early will avoid costly deployment failures and maintain trust with patients and staff. Teams should measure impact against specific operational metrics like procedure delays, supply chain friction, or clinician workload reduction. Professionals seeking to align their department's strategy with industry standards can explore an AI Learning Path for CIOs that covers governance, infrastructure security, and operational planning.
Your membership also unlocks:
