Dr. Skandhan
Amith Skandhan, MD, FACP, SFHM, a hospitalist and associate professor at UT Health in San Antonio, is so fascinated by the promise of artificial intelligence (AI) to transform medicine—especially hospital medicine—that he is completing advanced post-graduate training in AI and machine learning. His goal is to understand the technology deeply enough to bridge the gap between AI and medicine, advocate for its responsible deployment, and help redesign hospital workflows through a human-centered lens.
Dr. Skandhan said he believes hospitalists don’t need to get that involved in the technical side of AI. But they do need AI literacy, which means understanding the basics. “You have to learn when to rely on AI and when not to,” he said. “When it comes to AI, I feel what limits us is our imaginations, not the technology.”
The Future is at Hand
Dr. Skandhan encouraged his colleagues to imagine a near-future version of hospital medicine where AI quietly supports care behind the scenes. “I wake up, and AI has already been analyzing my patient list overnight. It highlights what’s likely to change with these patients, what needs follow-up, and what safety checks are required.”
On hospital rounds, AI synthesizes data while the hospitalist speaks with the patient. It forms answers to the patient’s questions, drafts notes in the background, and schedules needed clinical studies. “Can orders be placed without breaking our conversation?” he posed. “Can AI tell me, before sign-out, which patients are likely to deteriorate today, based on data and clinical trajectory? That technology exists—we just haven’t reimagined and integrated it.”
Dr. Skandhan said the best AI will be invisible, fully embedded into the electronic health record (EHR), and running quietly. If the system is working, he said clinicians simply show up and incorporate its insights into their care. “AI doesn’t replace our judgment, but it amplifies our clinical acumen. It gives back the time and space that we want to spend with our patients.”
But that future requires the right design and the right voices shaping it, he said. To get the version of AI doctors actually want and need— safe, empathetic, and practical for the doctor, for the hospital system, and, most importantly, for the patient—hospitalists must be part of its development, participating as the co-creators of the AI tools.
What Are We Talking About?
Dr. Kanjee
AI has been defined in terms of the simulation of human intelligence by computer systems that are able to absorb and analyze huge amounts of information. “AI has been around for a long time, but it very much entered the public consciousness recently,” especially since the launch of OpenAI’s ChatGPT in November 2022, said Zahir Kanjee, MD, MPH, a hospitalist at Beth Israel Deaconess Medical Center and assistant professor of medicine at Harvard Medical School, both in Boston.
“But we’ve been using artificial intelligence for a long time: for example, machine learning models. We’ve had sepsis alerts and all sorts of algorithms to help guide our thinking for a long time,” he said.
“I think when a lot of people think about AI right now, what they’re talking and dreaming and worrying about are the large language models, generative AI that can do a whole lot more than it used to do. We’re learning more, and every week there’s some new potential use case, or some other new thing that we see that they’re capable of doing that we didn’t think they were capable of doing,” Dr. Kanjee said.
“I’m absolutely loving the ways it can reduce some tedium in my job, the mindless tasks. What I’m currently using includes discharge summary tools and ambient listening technologies. The AI medical scribe we use is called Heidi.” He said AI can also help with diagnostic and management steps. “It could be potentially transformative for our jobs once we know exactly when and how to use it—or not.”1,2
Mr. Castillo
Ron Castillo, APRN, is the lead advanced practice practitioner at Yale New Haven Health and one of the medical informatics officers for Yale Health, both in New Haven, Conn. He has an integral role in introducing its AI scribe and documentation assistant and other AI tools to the inpatient setting. He also sits on the Hospital Medicine Steering Board for EHR manufacturer EPIC.
“As you can imagine, AI is a pretty hot topic on the EPIC board. One of the things we’ve talked about is the wide variety of perspectives and experiences with AI. Ask a clinician, administrator, or IT [information technology] person what AI means to them, and you will get different answers,” said Mr. Castillo.
How AI is encountered is partly a matter of what people’s interests are, and what they think is going to make their day easier or help them take care of their patients better, he said. Digital scribes, which listen to the doctor’s conversation with the patient and incorporate the pertinent findings into a draft note or discharge summary, are generating a lot of interest in hospital medicine right now.
What AI can do for hospitalists is also institution-dependent, with hospitals on a spectrum of where they are in adopting AI. “Some institutions are all in, and others are saying: Let’s take a step back and see what everyone else does, see what works and what doesn’t work, before we jump into this,” Mr. Castillo said.
“At Yale, we are starting to get our feet wet with documentation assistance. We have predictive models for things like clinical deterioration or readmission risk.” Yale Health uses clinical deterioration scores such as AgileMD’s eCART software, which continuously reviews nearly a hundred data points to determine the patient’s risk of clinical deterioration, he said. “The higher the score gets, we really should be thinking about what’s driving that and whether we need to revisit the patient’s care before we have a bad outcome on the floor,” he said.
“I think it’s one of those things, just like the EHR, that’s not going away. We have to figure out how we can use AI to our advantage, our patients’ advantage, to take good care of our patients and maintain that same quality and safety that we strived for pre-AI.”
Language and Large Language
Dr. Olson
AI is a huge, multi-faceted field, rapidly evolving, with major transformations measured in months, not years, said Andrew Olson, MD, FAAP, FACP, SFHM, an adult and pediatric hospitalist and division director of hospital medicine at the University of Minnesota Department of Medicine, in Minneapolis, Minn. Of the various applications of AI, those most relevant to medicine use large language models, defined as complex neural networks that can make predictions about the associations between words. They are trained on vast amounts of text data to understand and generate human-like language.
When AI tools first started entering the healthcare system, you had to actively engage with them, opening the tools on your computer, Dr. Olson said. “More and more, the tools are being incorporated into the EHR, and there are models you’re beginning to engage without even knowing that you’re engaging with AI.”
Dr. Olson has been an educational and clinical researcher throughout his medical career, with a particular interest in diagnostic error and how clinicians and teams make decisions. “I’m interested in these AI tools because I think they have an opportunity to change and improve how we make clinical decisions. But that needs to be studied.”2,3,4
He suggested that it’s possible to talk about AI’s role in hospital medicine in three big buckets. The first is using it in routine and mundane ways to help with the routine work that hospitalists do every day. “The sheer volume of tasks that need doing in hospital medicine is a challenge,” Dr. Olson said.
“So, can we use AI to help with tasks that maybe don’t require our whole cognition, like chart review or writing notes? I think discharge summaries will become a bigger part of AI for the practicing clinician, summarizing what happened in the patient’s course of hospitalization. We can do that now. They’re quite good.”
A second bucket is to use the technology for tasks that doctors were already doing, but to do them better. That might include displaying information that will enable the doctor to think better. And a final bucket is finding new and different ways to do things, or what an economist might call disruptive innovation, Dr. Olson said.
“For example, what we call a chart note, what if it became more like an individualized Wiki page for what you need to know about this particular patient, based on your specialty?” This could also apply to accessing clinical resources to answer patient care questions, moving from actively searching to having an AI tool suggest resources proactively.
“I think about other disruptive ways of communicating across handovers of care,” Dr. Olson said. “What if I’m working at night and I walk into a rapid response situation? Could I just talk to the computer and say, ‘Tell me what I need to know right now?’ I think we often get stuck in doing what we’ve always done, just wanting to do it faster and more easily. We need to think about how we might do things differently than we’ve ever done before.”
How Is AI Being Used?
Dr. Patel
Mihir H Patel, MD, MPH, MBA, FACP, CLHM, SFHM, is chair of the Inpatient Clinical Informatics Council at Ballad Health in Johnson City, Tenn. He recently relocated to Sacramento, Calif., where he practices hospital medicine part-time for Kaiser Permanente and Sutter Health. He also chairs SHM’s Health Information Technology Special Interest Group.
“Everyone’s talking about AI now, because it’s finally powerful enough to make a real difference, with greater capacity, broader capability, able to support more applications,” Dr. Patel said. “For hospitalists, that power sits right in our hands—through our phones, the EHR, and the tools we use at the bedside.”
He explained that AI can support risk prioritization, streamline hospitalist workflow, and strengthen safety across transitions of care. This includes automatically summarizing a hospital stay for discharge, highlighting readmission risk, and bringing attention to social determinants of health or other subtle risk factors that might otherwise be overlooked.
“AI feels like a second pair of eyes, continuously reviewing details in the background and surfacing what is clinically important, so that in the pace of hospital work I can stay present with the patient rather than with the screen,” Dr. Patel said. “Ambient documentation gives me back my time—turning pajama-time charting into family time, preserving not only hours but my passion for medicine.”
The promise of AI is that it is becoming an assistant that can make various aspects of the job easier, such as in the diagnosis and treatment of patients, said Peter Barish, MD, a hospitalist at the University of California, San Francisco (UCSF) in San Francisco. “Large language models are already great for assisting in diagnosis, but how, particularly on the treatment side, can AI help us adhere to evidence-based practice?” he said.
“I think there is a big question about how can this help us synthesize complete data to make diagnoses, reduce diagnostic error, and then choose appropriate management. Can it help us access information in ways that are easier to use, more patient-targeted, more streamlined?” Dr. Barish said.
UpToDate and the Open Evidence medical information platform have been widely used by physicians to look up quick answers to clinical questions, especially for less common conditions, and these are being enhanced with the advances in AI. Published clinical guidelines don’t always apply to every single patient, he said. “I think AI has the real potential to help customize and tailor evidence-based management to a particular patient.”
What Can Hospitalists Do?
Find your hospital’s AI Governance Committee and join it—or, if there isn’t one, help to form one, Dr. Skandan said. “Right now, we don’t have standardized AI governance in medicine, and the problem is that our AI policy will shape how we use it faster than the developments in AI technology itself.”
He also recommended seeking out further education, peer discussion, and shared learning. “Have conversations with other hospitalists through SHM and other forums.” A journal club or shared case discussions could be a great way to talk about what’s working and what isn’t.
One of the keys to working with AI is making sure it is guarding protected patient personal information, although that is also the responsibility of the hospital and its systems. The individual physician should not be entering protected data into software programs that aren’t under the hospital’s protected umbrella. But as more systems are integrated into the EHR, that will become less of an issue.
“We often talk about complacency in medicine and how that’s a bad thing,” Mr. Castillo said. “It’s no different with AI. If I just blindly sign my AI-drafted note and I don’t proofread it at all, and there’s false information or missing information in there, that’s on me.” The system needs to have people testing, looking at, and studying whatever the latest technology is.
“How do we make sure that AI doesn’t go off the rails? I think we continue to do what we always do as clinicians. We do our due diligence and our investigations. I think clinicians by nature are sort of skeptical beings to begin with, and that’s no different here,” Mr. Castillo said. “I’m certainly excited to be at the forefront of this technology in hospital medicine, but I want to make sure we do it right.”
Dr. Olson said it’s important for hospitalists to play with AI tools outside of their professional lives, especially the ones that are interesting to use—like a tool that can plug into their calendar and find open calendar slots for appointments.
Redefining Excellence
AI, obviously, is a huge, multifaceted phenomenon, bearing down on medicine and the rest of society like a massive tidal wave, with all sorts of implications, opportunities, and concerns. Robert Wachter, MD, of UCSF in San Francisco, a pioneer of the hospitalist movement in the U.S. and author of books about medical errors and the adoption of the EHR, has a new book coming out in February 2026 titled “A Giant Leap: How AI is Transforming Healthcare and What That Means for our Future.” It explores generative AI’s transformative potential for medicine as well as the lessons from the fraught implementation of EHRs, that digital transformation in healthcare is harder than it looks.
Dr. Skandhan noted that this huge topic of AI has generated a lot of excitement but also misunderstandings among physicians. “Personally, I feel this is a very exciting time for hospitalists,” he said, adding that fears that it is going to replace physicians seem unfounded. “Hospitalists stand at the crossroads of the hospital. Our voice is essential in determining how AI is used—for patients, for our teams, and for ourselves. We know better than anyone how it can be used safely, ethically, meaningfully.”
Dr. Skandhan believes hospitalists should be mapping their own workflow to identify where AI could help, and to understand the data any tool is trained on. “Bias isn’t just a technical concern. It’s a patient safety issue. AI can magnify disparities if we don’t put guardrails in place.” He also thinks SHM has a major opportunity to lead the way in national advocacy for responsible AI. “There must be transparency. There must be safety checks. Deployment should be clinician-led and hospitalist-led.”
Ultimately, he believes AI will raise the standard of care. “I think hospitalists who use AI will redefine excellence. It helps us process information faster, so we have more time to be empathetic, present, and human at the bedside—why we went into medicine in the first place.”
Larry Beresford is an Oakland, Calif.-based freelance medical journalist.
References
1. Kanjee Z, et al. Accuracy of a generative artificial intelligence model in a complex diagnostic challenge. JAMA. 2023;330(1):78-80. doi: 10.1001/jama.2023.8288.
2. Goh E, et al. Influence of a large language model on diagnostic reasoning: a randomized clinical vignette study. medRxiv [Preprint]. 2024:2024.03.12.24303785. doi: 10.1101/2024.03.12.24303785.
3. Goh E, et al. Large language model influence on diagnostic reasoning: a randomized clinical trial. JAMA Netw Open. 2024;7(10):e2440969. doi: 10.1001/jamanetworkopen.2024.40969.
4. Selvakumar S, Olson APJ. Faster and better for everyone-the future of AI-generated discharge summaries. J Hosp Med. 2025. doi: 10.1002/jhm.70186.