The Machine in the Room — What Physicians Actually Think About AI in Medicine

Artificial intelligence has arrived in medicine the way most disruptions do — not at the front door, but everywhere at once. It is in radiology reading rooms and psychiatric intake forms, in oncology decision support tools and obesity medicine protocols, in the ambient documentation software running quietly in the background while a physician tries to be present with a patient. The question that matters now is not whether AI will change medicine. It already is. The question is whether it will change medicine into something better.

MedStory Studio asked five physicians — across internal medicine, obesity medicine, OB-GYN, neuroscience, and oncology — to speak plainly about what they believe AI can do, what they are afraid it will do, and what those shaping the technology tend to miss. What emerged was not a consensus. It was something more valuable: a set of honest, specific, sometimes contradictory answers from people who spend their days in the room where medicine actually happens.

Part One: What Physicians Are Excited About

Giving Time Back

The single most consistent source of excitement across every physician we spoke with was neither diagnostic accuracy nor algorithmic sophistication. It was something simpler, and more fundamental: the possibility of getting time back. Time with patients. Time to think. Time not spent on the machinery of documentation, billing, and administrative throughput that has come to define so much of a physician's working day.

Shikha Jain, MD, FACP, an oncologist, names it directly — and frames it not as a productivity argument but as a human one:

"What excites me most is not AI replacing physicians. It is AI helping physicians be more present, more informed, and more connected to the patients in front of them. The goal should be to use technology to make medicine more human, not less."

— Shikha Jain, MD, FACP, Oncology

Elizabeth Garchar, MD, an OB-GYN and maternal-fetal medicine specialist, makes the same argument from the exam room floor. The administrative layer — charting, billing verification, documentation redundancy — is not where clinical training is meant to go. AI that absorbs that burden, she argues, is AI that directly reduces physician burnout and restores the cognitive space that good medicine requires.

Pattern Recognition at Scale

The second area of genuine excitement is the one that receives the most media attention — and, physicians say, the one where the hype is actually somewhat justified. AI is genuinely better than humans at certain kinds of pattern recognition: detecting anomalies across large imaging datasets, synthesizing research across thousands of papers, flagging risk in patient data that no single clinician could hold in mind simultaneously.

Russell Kennedy, MD, a neuroscientist specializing in anxiety and emotional dysregulation, frames this as an extension of the analytical mind — something that enhances rather than threatens physician cognition:

"AI represents a remarkable extension of the analytical mind. Used well, it can improve efficiency, reduce errors, and make high-quality medical knowledge more accessible to both doctors and patients."

— Russell Kennedy, MD, Neuroscience & Mental Health

In obesity medicine, Sejal Desai, MD, DABOM, points to a specific and underappreciated application: the identification of patterns humans miss. Not because physicians are inattentive, but because the human mind processes information sequentially, while AI can hold and cross-reference a patient's full history, lifestyle factors, genetic signals, and population-level data simultaneously. Early intervention, personalized treatment, and proactive outreach all become more achievable when the underlying pattern recognition is more powerful.

A New Division of Labor — and a More Human Clinician

Hillary Lin, MD, an internal medicine physician, offers the most structurally ambitious vision of what AI could mean for medicine — and it is not a vision of replacement. It is a vision of re-specialization. If AI can take over guideline recall, chart synthesis, triage coordination, and evidence comparison, then clinicians are freed to do what AI cannot: sit with patients in fear, help them navigate impossible decisions, and be fully present in the most vulnerable moments of a person's life.

"AI will be better at recalling guidelines, digesting massive charts, comparing evidence, and seeing patterns across complex data. We may not always be the smartest entity in the room anymore. But we will still be the only entity in the room that can truly care for the patient."

— Hillary Lin, MD, Internal Medicine

This is not a consolation prize for clinicians. It is, as Lin frames it, a reclamation — a return to the part of medicine that drew most physicians to the field in the first place, and that the current system has systematically crowded out. If AI can take over more of the machinery of medicine, clinicians may finally have the space to become more human again.

Part Two: What Physicians Are Worried About

Wasting the Moment

For all the optimism about AI's potential, the concern that recurs most urgently across these conversations is a specific kind of failure — not a dramatic collapse, but a quieter, more insidious one. The failure of using a transformative technology to make a broken system incrementally faster, rather than to build something fundamentally different.

"A lot of healthcare innovation is still being designed to sell into incumbents, and incumbents are largely motivated to win at the game of the status quo. That means AI may be used to make the current system faster, cheaper, and more scalable without asking whether the current system is worth preserving in the first place."

— Hillary Lin, MD, Internal Medicine

The risk Lin identifies is structural, not technical. The problem is not that AI is bad at medicine. The problem is that the institutions deploying it are optimizing for the wrong things. Bolting AI onto broken workflows produces faster broken workflows. The opportunity — the one that could actually change outcomes — is using AI's capabilities to reimagine what continuous, personalized, proactive care looks like.

Losing the Human Side

The concern most physicians name second is the one that carries the most clinical weight — particularly for those in specialties where the physician-patient relationship is not just an amenity but the mechanism of care itself. Kennedy, whose work centers on anxiety disorders and emotional dysregulation, makes the case with precision:

"Healing depends deeply on the quality of the relationship between clinician and patient. People begin to heal when they feel safe, seen, heard, understood, and genuinely cared for. Our nervous systems are wired for connection. AI may be able to simulate empathy with increasingly convincing language, but it does not actually feel."

— Russell Kennedy, MD, Neuroscience & Mental Health

Desai makes the same point from a different angle. The risk is not that AI will declare itself a physician. The risk is subtler — that overreliance on algorithmic outputs will erode the clinical intuition, lived experience, and contextual judgment that no dataset can fully capture. Medicine requires nuance. The most dangerous failure mode for AI in medicine is not a dramatic misdiagnosis. It is a slow erosion of the very faculties that make medicine more than pattern matching.

Bias, Equity, and the Tools Not Built for Everyone

Jain identifies what she considers the most serious systemic risk in AI's deployment across medicine — one that maps directly onto the existing failures of the healthcare system it is meant to improve. If AI is trained on biased data, or designed without the populations most in need of better care at the table, it will not merely fail to close existing disparities. It will encode and accelerate them.

This is not a theoretical concern. Medical datasets have historically overrepresented certain demographic groups and underrepresented others. AI trained on those datasets inherits those imbalances. Without deliberate corrective design — without diverse voices at the table, transparency about training data, and explicit commitment to equity outcomes — AI risks becoming another vector through which the system fails its most vulnerable patients.

Creativity, Clinical Reasoning, and the Risk of Automaticity

Garchar raises a concern that is less about equity and more about the epistemology of medicine itself. Medicine, at its best, is not a protocol execution. It is a creative act — the application of scientific knowledge to a specific, irreducible human being whose circumstances are never quite what the textbook described. The danger of AI, she argues, is not that it will be wrong. It is that it will encourage physicians to stop generating their own answers.

"It is really important to keep our creativity as we practice medicine and not just do things automatically because it's what we did last time."

— Elizabeth Garchar, MD, OB-GYN & Maternal-Fetal Medicine

In-place protocols have value. Standardization prevents certain categories of error. But clinical reasoning — the willingness to ask whether this patient, in this moment, requires something the protocol didn't anticipate — is not a redundancy to be optimized away. It is the core of what medicine is. Any deployment of AI that treats physician judgment as an obstacle rather than an asset has misunderstood both the technology and the profession.

The five physicians in this piece work across different specialties, different patient populations, and different corners of the healthcare system. They do not agree on the shape of AI's future in medicine. But they agree — clearly, and without ambiguity — on something more important: that the value of AI in medicine will be determined not by what the technology can do, but by what the people deploying it choose to build. The promise is real. The risk of wasting it is equally real. And the people best positioned to tell the difference — the ones in the room with patients — are the ones whose voices belong at the center of that conversation.