AI and the Hospitalist: Between Promise and Prudence
Artificial intelligence (AI) has quickly moved from novelty to near inevitability in hospital medicine. While tools like ChatGPT captured public attention in 2022, AI had already been shaping inpatient care long before, through predictive models, sepsis alerts, and risk stratification tools quietly embedded in our workflows. What has changed is not just the capability of AI, but its visibility and accessibility to frontline clinicians. It is therefore no surprise that a growing majority of physicians now see potential upside in its use.1
This issue highlights both the promise and the tension inherent in AI’s expanding role. On one hand, AI offers relief from some of hospital medicine’s most burdensome tasks: documentation, chart review, and synthesis of overwhelming amounts of data. Many hospitalists trained in an era when UpToDate revolutionized point-of-care learning are now watching trainees turn to AI-powered tools that surface primary literature and guidelines in seconds. Used thoughtfully, these tools can accelerate knowledge acquisition and support decision-making in complex cases.
Yet speed and efficiency come with trade-offs. As one article in this issue thoughtfully explores, ambient documentation and AI scribes may reduce time spent typing, but they also risk changing how we listen, how we distill information, and how we communicate with one another. Longer notes are not inherently better notes. The discipline of the one-liner, the curated assessment, and the abbreviated handoff remain central to safe hospital care. If AI amplifies noise rather than clarity, we risk solving one problem while creating another.
At a systems level, AI-driven predictive models have shown promise in identifying clinical deterioration and improving outcomes, including modest but meaningful reductions in mortality. These tools illustrate AI at its best: operating quietly in the background, augmenting, but not replacing, clinical judgment. However, AI is only as good as the data and incentives that shape it. Bias embedded in datasets, lack of personalization, and opaque decision-making processes should give hospitalists pause, particularly as AI begins to influence utilization management and post-acute care decisions. In fact, experiencing an insurance company denying patient care using AI tools is not uncommon anymore.
Perhaps the most important message from this issue is that AI is not something happening to hospitalists; it is something that must be shaped by them. Whether through governance committees, pilot programs, or daily use at the bedside, hospitalists have a responsibility to advocate for AI that is transparent, equitable, and clinically meaningful. AI may help us reclaim time, reduce burnout, and process information more efficiently. But preserving judgment, presence, and humanity in hospital medicine will always remain our work.
Dr. Mehta
Dr. Mehta is an associate professor of medicine and vice-chair of Inpatient Clinical Affairs at the University of Cincinnati in Cincinnati, Ohio. He is also the associate editor for The Hospitalist.
References
1. AMA Augmented Intelligence Research. AMA website. https://www.ama-assn.org/system/files/physician-ai-sentiment-report