The clinical reasoning pattern hasn't changed since Hippocrates. The vessels carrying it have, from paper charts to cloud EHRs. For fifty years, every vessel moved the data without understanding it. AI breaks that pattern for the first time. Not a better vessel. One that thinks.
This week I read three papers that made me happy. JAMA. NEJM. Nature Medicine. All randomized trials. All showing AI outperforming standard care. Then I read the methodology. None of them LLMs. The AI winning in top journals in 2026 was built before the hype cycle. The blueprint was always there.
A Nature study shows LLMs achieve 94.9% accuracy on benchmarks but only 34.5% when laypeople use LLMs on physician-created scenarios. The gap reveals something deeper: We measure the model in isolation. We deploy to a human in distress. The system fails at the intersection.
The physician's job has always been data synthesis. Fragments, approximations, family filters, descriptions of pill geometry. Consumer health AI was built for a different patient. One that has never walked through a clinic door.
Asimov gave robots three non-negotiable laws. Medicine gives physicians an oath. Healthcare AI has governance, but no runtime constitution. Until safety principles are enforced at the moment of output, not just in policy documents, we are deploying systems without the equivalent of “do no harm.”
A Nature Medicine study found ChatGPT Health under-triaged 52% of real emergencies. The deeper issue may not be the model, but its training data: AI learns from documented hospital records, yet it is deployed at first contact, where the most critical signals were never captured.
Governance by design is essential, until 3 a.m. In an emergency, a perfectly compliant blood bank workflow nearly cost a life. Healthcare AI doesn’t need less governance. It needs adaptive governance, rules that scale with risk and allow accountable overrides when seconds matter.
AI governance isn’t nested boxes or monthly committees. It’s architecture. Under HIPAA and GDPR, your “data steward” often can’t even review the data. Real governance is built into the plumbing, validation, lineage, and security enforced automatically. Otherwise, it’s theater.
Healthcare is rapidly adopting Customer Data Platforms to unify patient data and personalize engagement. Used as infrastructure, CDPs improve identity resolution, consent management, and outreach. But without clear governance, they risk optimizing for engagement over clinical need.
Digital twins are moving from factories to hospitals. We’re building replicas of patients, modeling disease in real time, and of physicians, capturing decades of expertise. The twin extends capacity while preserving oversight. What will happen when they no longer need us in between?
AI won’t replace physicians, but it will redefine them. As knowledge shifts from memory to machine, the xPhysician emerges: an AI-augmented generalist who combines deep expertise with end-to-end ownership, reshaping how medicine is practiced and learned.
“AI in healthcare” is no longer a category, it’s a confusion. We treat mature, low-risk systems the same as experimental, high-risk ones. The real issue isn’t AI itself, it’s matching governance to actual risk and use.