AI Resilience Report forHospitalists

Right now, AI in hospital medicine is mostly augmenting hospitalists — helping them work faster and reducing burnout — rather than replacing them. The biggest wave is "ambient AI scribes" that listen to patient encounters and draft notes automatically. Nearly two-thirds of hospitals using Epic have adopted ambient artificial intelligence (AI), with higher uptake among larger, not-for-profit hospitals and those with higher workload and stronger financial performance.

The American Hospital Association highlighted ^[1] real results: at Cleveland Clinic, ambient AI cut EHR note time by 14 minutes per day, and Mass General Brigham saw a 21.2% drop in burnout after using the tools.

For the highest-automation task — discharge summaries — Stanford Medicine piloted "MedAgentBrief" on a 24-bed hospitalist unit ^[2], where 11 physicians received AI-drafted summaries every morning; feedback showed only 2% hallucination rates and physicians reported significantly lower burnout, even though actual time saved was only about three minutes per summary. The Society of Hospital Medicine's publication The Hospitalist reports ^[3] that OpenEvidence, an LLM tool referencing medical literature, is now used by more than 40% of U.S. physicians for point-of-care decisions, though studies also flag "automation bias" — over-trusting AI suggestions — as a real risk.

Adoption is happening quickly on the paperwork side because the tools are commercially available (Abridge, Ambience, Microsoft Dragon Copilot) and the ROI is clear. Fierce Healthcare reported ^[4] that 75% of U.S. health systems now use at least one AI application, up from 59% in 2025, with more than half seeing at least a 2x return on investment. Clinical note-taking leads adoption at 68%.

But adoption for actual diagnosis and treatment decisions is much slower, and for good reasons: patient safety, legal liability, and the need for human judgment. The Stanford pilot ^[2] found 25% of AI-drafted summaries had omissions and 20% had inaccuracies — small errors humans must still catch. Hospitalist leaders emphasize that true clinical reasoning remains a human, contextual act ^[3], and hospitals are building evaluation frameworks before wider deployment.

So while AI will keep taking over documentation, the human skills of empathy, complex judgment, and coordinating care will keep hospitalists central to patient recovery for years to come.