Agentic AI in healthcare: Europe's opening to lead
Generative AI has settled into healthcare. It helps with documentation, summarization, and decision support. Agentic AI is a different move entirely: systems that make autonomous decisions and execute physical tasks in the real world.
That distinction matters. According to Paolo Denti, CEO and co-founder of Oversonic Robotics, Europe still has room to lead in agentic AI-even if US and Asian players dominate large language models and genAI.
Why this matters now
US companies like AWS and Nvidia have locked in deep partnerships with major medtech names, including Johnson & Johnson. Their genAI tools are already pushing drug discovery and predictive analytics forward.
Agentic AI sits closer to the bedside. As Denti puts it, "Agentic AI is about performing real physical tasks and for machines to make real decisions, not just to support healthcare practitioners with generic information as in the case of genAI." That's the opening Europe can take-because this segment is still developing.
Market signals healthcare leaders should watch
GlobalData projects AI in healthcare to reach a $19B valuation by 2027. GenAI adoption is high-one McKinsey study reports 85% of healthcare respondents already use it in some capacity. Agentic AI is earlier, but the potential is large: McKinsey estimates 70%-80% of medtech workflows and 75%-85% of pharma workflows could be advanced with agentic systems.
From concept to clinic: a working example
Oversonic's flagship product, RoBee, is an assistant robot powered by agentic AI. It's built to reduce pressure from staffing shortages in hospitals and other care settings-not to replace people, but to handle repeatable, labor-intensive tasks.
RoBee has been tested in neurorehabilitation wards at a major hospital in Rome. Denti notes the team saw stronger-than-expected acceptance from staff and patients. The robot supports rehabilitation by delivering higher volumes of therapy across neurological, cognitive, motor, and language exercises.
Oversonic is applying for Class IIa medical device certification in Europe. Once complete, the robot will be commercially available. For context on EU device pathways, see Regulation (EU) 2017/745 (MDR) (official text).
Practical steps for healthcare providers exploring agentic AI
- Start where physical throughput limits care: rehabilitation sessions, patient transport, supply runs, routine vitals, basic bedside assistance.
- Define clear guardrails: which decisions the system can make, what requires human sign-off, and how overrides work.
- Integrate with existing workflows and data systems early to avoid creating a parallel process that staff won't adopt.
- Run controlled pilots with measurable endpoints: staff time saved, therapy minutes delivered, patient outcomes, safety incidents, and satisfaction.
- Plan for compliance and liability: MDR classification, risk management, logging, cybersecurity, and incident response.
- Invest in training and change management so clinicians trust the system and know exactly how to work with it.
Europe's path forward
The genAI race moved fast and big players set the pace. Agentic AI is more open. Europe can win here by focusing on clinically validated systems that do real work on the ward, backed by rigorous certification and evidence.
If you're mapping skills and training for your team, you can scan healthcare-relevant AI courses by job role here: Complete AI Training - Courses by Job.
Bottom line: Agentic AI won't replace clinicians. It will expand what your teams can deliver-safely, consistently, and at scale-if you choose high-evidence use cases and implement with discipline.
Your membership also unlocks:
