Agentic AI could coordinate care, retrieve information, and automate routine tasks, helping health systems cope with growing staff shortages and increasingly complex workflows. Yet healthcare remains reluctant to cede meaningful autonomy to AI systems. The industry is often more comfortable with familiar inefficiencies and human error than with the possibility of an unfamiliar machine error.
At first glance, healthcare appears to be an ideal environment for AI agents. Few industries face greater pressure from workforce shortages, rising demand and expectations, and increasing complexity. But the closer AI agents move toward decisions that affect patient care, the more cautious healthcare leaders become. "Autonomous" seems to be a forbidden word that stirs up fears of the dehumanization of healthcare and accountability.
Medicine is like a minefield. Risks can explode unexpectedly
It’s obvious that the application of AI agents in healthcare is far different from that in other industries. An autonomous system that incorrectly categorizes an email or schedules the wrong meeting creates inconvenience. An autonomous system that contributes to a missed diagnosis or an inappropriate treatment recommendation creates an entirely different category of risk.
While there is broad agreement that agentic AI could assume greater responsibility for routine operational tasks, there is considerably less enthusiasm for granting similar autonomy in diagnosis or treatment. However, healthcare does not even trust autonomous clinical decision-making. The technology continues to evolve rapidly, yet uncertainty remains regarding reliability, accountability, and the consequences of failure.
The challenge extends beyond technological performance. As AI systems become more capable, clinicians may become increasingly inclined to trust their recommendations. This phenomenon, often described as automation bias, introduces a new layer of complexity. The danger also includes humans becoming insufficiently critical of its conclusions. The result is a healthcare sector that remains open to innovation while maintaining a cautious approach to autonomy. Progress is welcomed. Responsibility remains non-negotiable.
Administrative, repetitive work could be automated, but the data is messy
Healthcare professionals spend substantial portions of their working day documenting encounters, searching for information, navigating fragmented systems, and completing administrative processes. These activities consume time, contribute to burnout, and reduce the time available for patient care – it’s estimated that administrative tasks consume up to 50% of a clinician's workday.
Agentic AI offers a compelling solution precisely because many of these tasks are structured, repetitive, and associated with relatively low levels of clinical risk. Scheduling appointments, coordinating referrals, retrieving information, managing documentation, and supporting operational workflows are areas where intelligent agents could deliver meaningful efficiency improvements.
The significance of such improvements should not be underestimated. Healthcare systems across the world are struggling with shortages of physicians and nurses while simultaneously facing increasing demand from aging populations. Every hour returned to clinical staff has the potential to improve patient care, reduce waiting times, and alleviate pressure on overstretched organizations.
But even university medical centers open to innovation often face problems that slow the adoption of AI, such as a lack of robust infrastructure. Hospitals remain burdened by fragmented data environments, disconnected information systems, and inconsistent data quality. An intelligent agent can only perform as well as the information available to it. Governance frameworks, auditability, interoperability, and trustworthy data architectures have become prerequisites for broader deployment.
Patients will be the strongest advocates for AI agents
While healthcare organizations remain focused on governance, regulation, and implementation challenges, many patients are already generating vast amounts of health-related data and searching for better ways to make sense of it.
The traditional healthcare model was built around relatively limited datasets collected during occasional encounters with clinicians. But now, wearable devices, home monitoring technologies, genetic testing, digital therapeutics, and patient-reported outcomes are producing a continuous stream of information that extends far beyond the walls of hospitals and clinics.
Yet the value of this information often remains unrealized. Clinical teams operate under significant time constraints, while health systems continue to struggle with fragmented records and disconnected data sources. Patients often accumulate hundreds of pages of medical documentation over the course of their treatment journeys. Extracting meaningful insights from such volumes of information presents a challenge even for experienced clinicians.
This is where AI agents could prove particularly valuable. Their greatest contribution may lie in helping patients and healthcare professionals navigate an increasingly complex information landscape. Agents could consolidate records from multiple sources, identify relevant trends, highlight missing information, and prepare summaries tailored to specific clinical contexts. Patients already use AI extensively, whether asking ChatGPT for simple advice on daily decisions or uploading lab test results to better understand their health.
However, there are clear boundaries around patient acceptance of automation. There is considerable enthusiasm for tools that improve access to information, simplify administrative processes, and support care coordination. Far less enthusiasm is observed for technologies that attempt to replace human relationships. This might soon change – trust in the technology isn't set in stone, and it’s evolving alongside the benefits it offers. Recent studies suggest that trust in AI is decreasing; meanwhile, the health-related use of AI chatbots is rising.
Coordination of care might be a perfect niche for agentic AI
So what’s the future of agentic AI in medicine? Rather than creating autonomous digital physicians, healthcare organizations are likely to deploy networks of specialized agents that operate across different stages of the care journey. One agent may collect patient information before an appointment. Another may retrieve records from multiple systems. A third may prepare summaries for clinicians or coordinate follow-up activities. The cumulative effect could be substantial even though each individual function remains relatively narrow.
Such an approach aligns with the broader direction of healthcare transformation. The greatest inefficiencies often arise not from a lack of clinical expertise but from poor coordination, fragmented information, and administrative complexity.
Sigrid Berge van Rooijen, a health AI consultant (Netherlands), describes future scenarios in which AI agents support patients throughout their healthcare journeys by coordinating appointments, organizing information, and facilitating communication between stakeholders. These activities require intelligence and initiative, yet they stop short of replacing clinical judgment.
Building an agent capable of making clinical decisions is increasingly a technical challenge. Yet medicine rarely offers the certainty that technology thrives on. Clinical decisions are often made with incomplete information, competing risks, and evolving circumstances. Physicians draw not only on data, but also on observation, experience, communication, and an understanding of the broader context surrounding a patient.
For that reason, AI agents are likely to make their greatest impact in care coordination. Organizing information, connecting fragmented systems, summarizing records, and guiding patients through increasingly complex care pathways are tasks that align naturally with the strengths of agentic AI.
There will inevitably be cases where AI agents make mistakes, cause harm, and reignite debates about whether medicine and artificial intelligence are truly compatible. Yet healthcare's challenges are too significant to ignore the potential benefits. Faced with growing workforce shortages, rising demand, and overwhelming volumes of data, health systems can scarcely afford to miss the opportunity that agentic AI presents.
