Episode 30 · Sep 20, 2026 · A Healthy Conversation
with Jeff & Lisa
About this episode
Dr. Lisa Belisle and Dr. Jeff Barkin open by taking stock of how artificial intelligence has already entered clinical practice, well ahead of public awareness. Barkin describes using ambient AI scribes, including through the HIPAA-compliant telehealth platform Doxy.me, to generate accurate, condensed progress notes in real time, freeing clinicians to maintain eye contact and attend fully to patients. Belisle adds context on the scale of physician adoption and frames the central tension: tools that reduce documentation burden can also reproduce the same distancing effects that made electronic health records a source of clinician burnout in the first place. Both co-hosts argue that clinicians must be active participants in shaping how AI enters care settings, not passive recipients of systems designed without them.
The conversation turns to specific applications and their complications. Barkin discusses using a custom AI tool to streamline complex memory-loss evaluations, and both co-hosts describe using OpenEvidence, a free clinical-reference app that grounds answers in peer-reviewed literature to reduce the risk of fabricated citations. They also explore patient-facing uses, including Google's NotebookLM, which Barkin has introduced directly to patients as a way to explain medical information from trusted sources without generating unsupported content. Against these benefits, the co-hosts weigh serious concerns: AI bias baked into training data, illustrated by pulse oximeters that perform differently across skin tones and a birth-outcome calculator that produced disparate results for Black and Hispanic women; data privacy risks when clinicians copy identifiable patient details into outside tools; the discovery that AI systems analyzing mammograms can predict cardiovascular risk in ways their designers did not anticipate; and the broader cybersecurity exposure that advanced AI models create for hospitals and critical infrastructure. The 2025 episode in which Swedish Prime Minister Ulf Kristersson publicly acknowledged using ChatGPT for policy guidance serves as their opening illustration of how AI governance questions have moved from technical settings into public life.
The final segment addresses governance directly, describing it not as a policy document on a shelf but as a living process: committees of clinicians, privacy officers, lawyers, and operations staff deciding which tools may enter a health system, how patient information is protected, and how problems get reported. Belisle and Barkin connect this to a broader concern about skill atrophy, drawing a parallel between over-reliance on GPS eroding map-reading ability and over-reliance on AI potentially eroding independent clinical reasoning. They argue the answer is balance rather than avoidance: using capable tools deliberately while preserving the human judgment, critical thinking, and direct attention to patients that no model can supply. The episode closes with the observation that getting clinicians and patients to trust a tool, use it well, and remain alert to its limits is itself the work, and that it is fundamentally a human challenge.
From this episode
“This is where we as humans have to govern a machine, a machine that we don't even know how it thinks.”
“Good governance is mindful of identification and avoids identifying factors. So I think it can be done. I think we can have our cake and eat it too.”
“If the medical record data is itself reflective of our beliefs about people, then the artificial intelligence is merely going to reflect that back.”
“You have a committee of people, typically clinicians, lawyers, privacy and operations staff, who are deciding which tools are even allowed to come into the health system, and once they're in there, who uses them and how a problem gets reported.”
“Getting people to trust a tool and use it well is the actual work, and that's a human issue. It's not an AI issue, it's not a software issue, it's not a governance committee issue. It's a human issue.”
“When people hear the word governance, they think about a policy on a shelf, but inside a healthy system, it's so much more alive than that.”