Practical AI in Healthcare

Every technology healthcare has adopted moved through the same curve: a bleeding-edge few went first and documented what broke, and the majority followed. Large language models skipped that entirely. Jeremy Harper, a biomedical informatician who has worked at Epic, Ohio State, and Regenstrief, and who wrote Large Language Models (LLMs) for Healthcare, explains what we gave up by going all at once. Ambient scribes are everywhere, an...

Chris Dymek came to healthcare informatics by way of philosophy, and she still thinks like a philosopher: the first question is what we actually mean. As Director of Digital Healthcare Research at AHRQ, she funded AI research, helped launch work on AI and patient safety, and wrote a request for information asking healthcare organizations how they thought about AI literacy. It was never published. She left in May 2025 and carried th...

The internet is full of wellness advice built on a single cherry-picked study. Renee Deehan, a molecular and cell biologist who leads science and AI at InsideTracker, spent two decades building the opposite. In this episode, she explains why the core of their recommendation engine is symbolic AI, knowledge representation and reasoning, rather than a large language model: it's deterministic, fully auditable, and by design cannot hal...

Dr. Vimla Patel has spent four decades studying how physicians actually reason, and what happens when technology ignores it. In this episode, she explains the difference between forward reasoning (the fast, pattern-driven hallmark of expertise) and backward reasoning (the slower, hypothesis-testing mode of novices), and why most clinical AI is built for the wrong one. Through two vivid cases, a textbook expert diagnosis and a near-...

Peter Embi has spent his career at the intersection of medicine and informatics: coining the term "algorithmovigilance," serving as the nation's first Chief Research Information Officer, and leading AI work at Vanderbilt. He is also a patient. A rare adrenal tumor took nearly 15 years and a self-diagnosis to catch, and it nearly killed him. In this conversation, Embi connects that diagnostic odyssey to the case for monitoring clini...

Interoperability has been healthcare's twenty-year broken promise, but something genuinely changed. Mika Newton, CEO of xCures, explains how provider data exchange jumped from roughly 30% to 85–90% in just two years, and why that's only half the story. Moving records, it turns out, was never the hard part. As Newton puts it, "records travel, but they don't translate" — most of a record arrives as duplicated notes and sc...

In their sixth Reflections episode, Steve and Leon look back across five conversations (Sarah Rossetti, Jeff Smith of ONC, Hugo Campos, Fred Bennett, and Zak Kohane) and decide, deliberately, not to force a single grand theme. What surfaces anyway is one question asked five ways: as AI gets genuinely capable, what still has to be human, and what makes it trustworthy? They cover Rossetti's CONCERN system that models the nurse instea...

A quiet war is underway over who controls your health data in the age of AI, and the front-runners aren't the usual suspects. Dr. Isaac "Zak" Kohane, Chair of Biomedical Informatics at Harvard Medical School and founding editor of NEJM AI, traces the line from SMART on FHIR (the accidental standard he helped create in 2009) to today's battle for the doctor-facing AI layer, where companies like OpenEvidence are outrunning the EHR in...

For its first-ever live episode, recorded before an audience at New York Tech Week, Practical AI in Healthcare sits down with Fred Bennett, founder and CEO of PatientTalker — an ambient-AI app built for the patient rather than the clinician. (Steve Labkoff is a disclosed advisor to the company.) Bennett traces the idea to his father's cardiology visit, where three family members left with three different memories of the same ...

Patient advocate Hugo Campos spent more than a decade fighting for access to the data from his own implanted defibrillator. When the system wouldn't budge, he stopped trying to reform it and started building around it. In this episode, Hugo shows how he used agentic AI coding tools to create OpenKP, an open-source app that liberates his records from inside Kaiser Permanente, despite calling himself a non-coder. He and the hosts unp...

What happens when the rules for getting AI into clinical care are written by someone who has spent his career inside both the advocacy world and the government? In this episode, we talk with Jeff Smith of ONC at HHS, the first government official on Practical AI in Healthcare. Smith walks us through ONC's proposed HTI-5 rule, including a striking move to treat AI agents as "users" with the same data-access rights as clinicians, and...

On National Nurses Day, Practical AI in Healthcare welcomes its first nurse: Sarah Rossetti, RN, PhD, of Columbia University. Her CONCERN early warning system takes an unusual approach to predicting patient deterioration. Instead of modeling a patient's vital signs and labs, it models the nurse's documentation behavior, since the frequency and timing of charting reflect clinical concern long before the numbers move. In a 74-unit ra...

In their fifth Reflections episode, Steve and Leon look back across six conversations (Matt Truppo at Sanofi, Ted Shortliffe, Barry Chaiken, David Hidalgo-Gato, and Danny van Leeuwen) to ask a sharper question: how specialized does AI have to be to actually work? The throughline is depth. The LLM is a commodity, and so, increasingly, is the generalist agent. What stays scarce is specialization in a workflow, the revival of symbolic...

Danny van Leeuwen is a nurse of 50 years, a multiple sclerosis patient, host of the Health Hats podcast, and a serial member of national outcomes panels at CMS, AHRQ, PCORI, and the National Academy of Medicine. He has tried more than 100 health apps and uses five. In this episode, he explains how he uses AI to interrogate his own chart, surface symptom patterns, and prepare for clinical encounters, and he shares his Three T's and ...

David Hidalgo-Gato is the founder and CEO of Cleo Health. While more than 100 competitors were building generic ambient AI scribes, David's team chose emergency medicine and stayed with one design partner for nine months and roughly 50 product iterations before launching. The result: an average 54-minute time savings per shift, a patient-assignment tool that turned a four-hour process into 15 to 20 minutes, and use across 400+ hosp...

When physician Barry Chaiken was diagnosed with prostate cancer, his clinical training gave way to fear. It took a friend asking, "What are you doing?" to snap him back into doctor-mode thinking. That experience reshaped how he sees AI in healthcare. In this episode, Chaiken draws on his dual perspective as physician and two-time cancer survivor to argue that consumer health AI is failing patients, not because the models are bad, b...

In Part 2 of our conversation with Matt Truppo, Global Head of Research Platforms and Computational R&D at Sanofi, we move from discovery to development, where the real stakes begin. Matt unpacks the promise and limitations of “digital patient twins,” a concept often described as the holy grail of drug development. With nearly 90% of drugs failing in clinical trials, even modest gains in predicting efficacy or patie...

Ted Shortliffe built MYCIN at Stanford in the 1970s, one of the first medical AI systems ever deployed in a clinical setting. Five decades later, he joins Steve and Leon to examine what has persisted in clinical decision support — above all, the demand for explainability — what has changed (computational power finally caught up to the ideas), and what the field may have lost along the way. The conversation includes a di...

AI in drug discovery has been long on promise and short on delivery. Matt Truppo, Global Head of Research Platforms and Computational R&D at Sanofi, presents a different picture. His team used AI to identify 10+ novel drug targets in 12 months, screen 30 million target combinations in days, and produce AI-designed compounds with 75% synthesizability. But Truppo is equally candid about the gaps: data integration, explainability,...