As artificial intelligence systems grow more capable of handling medical knowledge and decision-support tasks, the healthcare profession faces an unexpected crisis: redefining what it means to be an expert clinician. The rapid advancement of AI in medicine is unfolding at the same moment clinicians are drowning in data, forcing a fundamental rethinking of clinical training, expertise, and the relationship between doctors and machines.

What Is Changing About How Doctors Work With AI?

The evolution of AI in healthcare has moved far beyond simple diagnostic assistance. Early machine learning systems were trained on labeled data to perform narrow tasks, such as identifying abnormalities in medical images. Today's AI systems, built on transformer technology, can learn from data without human supervision and generate complex outputs. Some systems now operate as "agentic frameworks," where multiple AI agents work together to perform distinct tasks simultaneously.

This shift is creating what researchers describe as an "epistemic shift" in medicine, one that touches how clinicians process information, how patients seek it, and fundamentally what the profession means by expertise. Pete Clardy, MD, Director of Clinical Enterprise at Google for Health, explained this transformation at Cleveland Clinic's AI Summit for Healthcare Professionals, noting that "AI is in its infancy. It's as bad as it will ever be right now, and the rate of change is remarkable".

Pete Clardy, MD, Director of Clinical Enterprise at Google for Health

"We find ourselves collectively in this situation of too much data, not enough information," said Dr. Clardy, describing the core challenge facing modern clinicians.

Pete Clardy, MD, Director of Clinical Enterprise at Google for Health

The role of AI is shifting from automating narrow tasks to augmenting clinicians' ability to make sense of increasingly complex medical information. Rather than replacing clinical decision-making, AI is helping clinicians "find signal from the noise, organize and summarize complex information," according to Dr. Clardy.

Dr. Clardy

How Are AI Systems Performing in Real Clinical Settings?

Research is now testing AI systems in increasingly sophisticated clinical roles. In a single-center study cited by Dr. Clardy, a text-based AI system obtained patient histories, incorporated information from medical records, and generated clinical documentation along with assessment and plans. Notably, patient trust increased after interacting with the AI system, and the AI-generated differential diagnoses and management plans were similar in quality to those created by human clinicians.

The research has expanded to include video-based interactions, where two AI agents work in parallel: a "talker" agent that maintains conversational flow and a "planner" agent that monitors video and conversation for areas needing clarification. The AI co-clinician performed near human levels in triage and history-taking and outperformed comparison models in clinical reasoning. However, the research also revealed an interesting gap: AI performed better than humans in certain empathy-related tasks, while humans outperformed video-based AI in counseling and communication.

What Are the Hidden Risks of AI Integration in Medical Training?

As AI becomes embedded in clinical workflows, a new concern has emerged: the risk of "deskilling," "misskilling," and what Dr. Clardy calls "never skilling." This last term describes a troubling scenario where medical trainees who begin their education in an AI-enabled environment may never develop the ability to practice independently without AI assistance.

The challenge is particularly acute because how clinicians manage AI integration depends on where they are in their developmental curve. A senior physician with decades of experience can use AI as a tool to augment their existing expertise. A medical student, by contrast, may become dependent on AI before developing foundational clinical reasoning skills.

How Should Healthcare Organizations Implement AI Successfully?

Dr. Clardy emphasized that technological sophistication alone does not determine whether organizations successfully implement AI. Instead, successful implementation depends on several critical factors:

• Change Management: Organizations must actively manage how workflows and roles change when AI is introduced, not simply deploy the technology and expect adoption.
• Stakeholder Alignment: All relevant parties, from clinicians to administrators to patients, must understand and agree on how AI will be used.
• Problem Clarity: Organizations must be "crystal clear" about the specific problem to be solved or goal to be achieved, whether that is automating a process, augmenting human work, or pursuing innovation.
• Low-Risk Starting Points: Implementation should begin with low-risk use cases before expanding to higher-stakes clinical decisions.

"This is where we fail more often than on the basis of technology," Dr. Clardy stated. "We are insufficiently crisp on the problem to be solved".

Dr. Clardy

What Does Clinical Expertise Mean in an AI-Enabled Era?

Perhaps the most pressing question facing medical education and practice is definitional: what should expertise look like when AI can memorize vast amounts of medical knowledge, organize it, and understand its contextual relevance? Medical training has historically taught clinicians to become strong pattern recognizers, but as the complexity of patient presentations has grown, so too has the pattern-recognition task.

"I don't have an answer," said Dr. Clardy when asked what expertise should look like in an AI-enabled environment. "But I do think that this is going to be one of the most important questions that we address going forward."

Pete Clardy, MD, Director of Clinical Enterprise at Google for Health

The question is not merely academic. How the healthcare profession answers it will shape medical education, clinical practice standards, and the relationship between clinicians and AI systems for decades to come.

How Is the Patient-Clinician Relationship Evolving?

Another significant trend is how quickly patients are beginning to use AI independently to understand their own health information, both alongside and separately from clinicians. This is creating what Dr. Clardy describes as an evolving "triadic relationship" between clinician, patient, and AI.

Dr. Clardy

These tools may help level the historic information asymmetry between patients and clinicians, allowing patients to review their medical records in detail and arrive at appointments with more informed questions. Dr. Clardy connected this rapid adoption to a lesson from a National Academy of Medicine discussion about trust and innovation in healthcare, during which a patient advocate remarked that "innovation in health care moves at the speed of desperation".

The implications are significant. As patients increasingly use AI to navigate their own health, clinicians must adapt to a new dynamic where patients arrive with AI-generated insights and questions. This shift underscores why defining clinical expertise in the AI era is so urgent; clinicians will need to know not just medicine, but how to work effectively alongside both AI systems and increasingly informed patients.

What Does This Mean for the Future of Medical Practice?

Dr. Clardy described the current period as an important "tool shaping moment," noting that AI tools remain highly adaptable but will "harden over time." This means the decisions made now about how AI is integrated into clinical practice will have lasting consequences. He invoked Father John Culkin's observation: "We shape our tools, and therefore our tools shape us".

Dr. Clardy

The healthcare profession stands at a crossroads. The technology is advancing rapidly, patients are adopting AI tools at an accelerating pace, and clinicians are being asked to integrate AI into their workflows without clear guidance on what expertise should look like in this new environment. The answers to these questions will determine not just how AI is used in medicine, but what it means to be a doctor in the decades ahead.