Did you know that the average physician spends hours each day on documentation? It can be debilitating and frustrating. It can keep you from focusing on the actual practice of medicine and the patients sitting right in front of you. Millions of doctors suffer from charting fatigue every single year. In some cases, the administrative burden can be so severe that it limits a person’s ability to function normally in their clinical role. If you are experiencing this kind of burnout, you are not alone. Electronic health record documentation is one of the most common types of strain people experience in modern healthcare.
For a while now, passive ambient scribes have offered some relief. However, the technology is taking a massive leap forward right now. Hospitals are beginning to introduce active clinical AI agents. These systems do not just listen and transcribe your conversations. They actively interface with patient data, prepare charts in advance, and propose clinical actions for your review. In this blog post, we will discuss how Penn Medicine and K Health are deploying clinical AI agents, what the research says about their real-world impact, and when this technology should drive your workflow decisions.
What exactly are clinical AI agents?
To understand the shift, you have to look at what ambient AI scribes currently do. Standard scribes simply convert spoken conversations into clinical notes. Clinical AI agents go much further. They are active participants in the clinical workflow. These agents pull relevant data from the electronic health record, summarize long patient histories, and draft differential diagnoses before the patient even enters the room.
Let’s take a look at the recent development of general-purpose biomedical systems to see how this works. Huang K et al, *bioRxiv : the preprint server for biology* 2025, developed Biomni as a general-purpose biomedical AI agent capable of managing complex reasoning tasks. It interacts with external tools to gather information before proposing an action. This represents a massive shift from a passive dictation tool to an active chart preparer. Your clinical AI agents will soon handle the initial chart review and summarize years of clinical data.
The complete autonomous workflow
The workflow changes entirely when the AI has agency. Instead of waiting for you to dictate a plan, the agent reviews the patient’s incoming symptoms and matches them against the medical history. It then writes up a proposed plan. This gives you a starting point. It helps treat various conditions by ensuring no historical data point is missed. Moreover, it allows the physician to step into the role of an editor rather than a data entry clerk.
How Penn Medicine is pushing beyond ambient scribes
How is Penn Medicine pushing beyond ambient scribes? The recent partnership between Penn Medicine and K Health marks a very specific turning point in healthcare technology. They are deploying clinical AI agents across their entire health system to actively prep charts and suggest orders. This deployment represents a major milestone in academic medical centers trusting autonomous AI workflows over traditional administrative software.
The system looks at the patient profile before the visit. It asks the patient pre-visit questions through a chat interface. It then synthesizes all that data into the medical record. It proposes clinical actions for physician review. Yes, you heard that right. The AI drafts the medical plan, and the physician simply reviews it, edits it, and signs off.
Role of clinical AI agents in large health systems
Integrating these agents into a massive academic center requires careful planning. Your doctor will likely start seeing AI-generated suggestions pop up in the chart before the encounter begins. The Penn Medicine model shows that large institutions believe the technology is safe enough to draft clinical orders. However, the physician always remains the final decision maker. The AI does not finalize the chart or send prescriptions to the pharmacy on its own. It simply removes the friction of creating the initial draft. Besides this, the agents can identify gaps in care such as missing preventative screenings or overdue lab work.
How clinical AI agents interface with the EHR
Let’s take a look at how clinical AI agents actually connect with the electronic health record. The biggest bottleneck in healthcare technology has always been interoperability. Standard software often struggles to pull data from major platforms seamlessly. However, modern clinical AI agents are designed to sit directly on top of these databases. They read the structured data such as lab results and vital signs. They also read the unstructured data such as old consultation notes and discharge summaries.
When a complex patient comes into the clinic, the AI can instantly summarize a decade of cardiology notes. It triggers the altered motor control group or whatever specific pathways the hospital has defined. This means your doctor does not have to spend twenty minutes scrolling through old encounters just to figure out why a specific medication was stopped five years ago.
The impact on pre-visit planning
Pre-visit planning is essential for a smooth clinic day. In many practices, medical assistants spend hours the day before gathering records and prepping charts. Clinical AI agents automate this entire process. They generate a concise briefing document for the physician. This document highlights abnormal labs, flags missing preventative care, and suggests a preliminary plan for the upcoming visit. This creates an all-rounded solution for clinics struggling with staffing shortages and high patient volumes.
The current reality of the documentation burden
Is the documentation burden really that bad? The short answer is yes. Everyone wants to know if these tools actually save time and reduce stress. Let’s look at the data on ambient AI scribes, which set the baseline for what active clinical AI agents will achieve. Shah SJ et al, *Journal of the American Medical Informatics Association : JAMIA* 2025, found that ambient artificial intelligence scribes significantly affect physician burnout and perspectives on usability. The documentation burden is a primary driver of burnout across almost every medical specialty.
Ma SP et al, *Journal of the American Medical Informatics Association : JAMIA* 2025, looked closely at the utilization and impact on documentation time. When clinicians use these tools, the time spent writing notes drops dramatically. Hassan H et al, *Applied clinical informatics* 2025, conducted a systematic review of the clinical implementation of artificial intelligence scribes in health care. They found consistent reductions in charting time across multiple specialties ranging from primary care to surgical subspecialties.
Clinician experiences with new technology
Duggan MJ et al, *JAMA network open* 2025, reported on clinician experiences with ambient scribe technology to assist with documentation burden and efficiency. The relief is palpable for many doctors. They can finally look the patient in the eye instead of staring at a computer screen. Shah SJ et al, *JAMA network open* 2025, evaluated physician perspectives on ambient AI scribes and noted that while the initial burden drops, the systems are certainly not perfect yet. Sometimes the AI misses clinical nuance or formats the note poorly. However, moving from passive scribes to active clinical AI agents that pull directly from the EHR will likely bridge this gap by grounding the AI in the patient’s actual medical history.
Will agentic AI change radiology and oncology?
Will agentic AI change radiology and oncology? Let’s look at some examples in highly specialized fields to see where this is going. Khosravi B et al, *Radiology. Artificial intelligence* 2026, described the evolution of agentic AI in radiology from large language models to future clinical integration. Radiology is an essential field for this technology because of the massive volume of imaging data. Clinical AI agents can pre-read scans, compare them to historical imaging, and draft a preliminary report for the radiologist to review.
In the cancer space, Truhn D et al, *Nature reviews. Cancer* 2026, highlighted how artificial intelligence agents operate in cancer research and oncology. These agents can sift through genomic data, pathology reports, and imaging to suggest targeted therapies. Guo Y et al, *Advanced science (Weinheim, Baden-Wurttemberg, Germany)* 2025, explored radiogenomics based on artificial intelligence to achieve non-invasive precision medicine in cancer patients. By tying images directly to genetic profiles, the AI provides a comprehensive picture of the disease state.
Self-driving laboratories and automated research
The technology is even pushing into physical laboratory spaces. Liu S et al, *Biofabrication* 2026, touched on self-driving bioprinting laboratories. In these setups, the AI agent controls physical equipment to run experiments and analyze results autonomously. The applications are clearly expanding beyond simple text generation. Your oncology department could soon rely on clinical AI agents to cross-reference a patient’s tumor genetics with the latest clinical trials instantly.
Adoption of autonomous AI in diverse practice types
Adoption of autonomous AI in diverse practice types is another major area of interest. You might think this technology is only for massive academic centers like Penn Medicine. However, that is not entirely true. Goldstein J et al, *Frontiers in digital health* 2023, looked at the determinants for scalable adoption of autonomous AI in the detection of diabetic eye disease in diverse practice types. They collected real-world data to find key best practices. They discovered that smaller clinics can also integrate these tools successfully if they map out their workflows carefully and train their staff properly.
Kazemzadeh K, *Medical hypothesis, discovery & innovation ophthalmology journal* 2025, discussed artificial intelligence in ophthalmology, pointing out the opportunities and challenges of bringing these tools into daily practice. Whether you run a solo practice or work in a large health system, clinical AI agents will eventually become an essential part of your toolkit.
Implementation outside of primary care
Román-Belmonte JM et al, *Frontiers in bioscience (Landmark edition)* 2021, noted similar trends in musculoskeletal conditions. Your physical therapy or orthopedic clinic could use these agents to track patient recovery trajectories and suggest protocol adjustments. The AI can review progress notes from physical therapy sessions and alert the referring surgeon if the patient is falling behind expected milestones. This approach does not require the physician to manually review every single chart every day.
What are the limitations and ethical challenges?
What are the limitations and ethical challenges? This is where we have to pause and look at the counter-evidence. Although it may seem counterintuitive given all the hype, there are significant hurdles to overcome before these systems run completely independently. Hager P et al, *Nature medicine* 2024, evaluated the limitations of large language models in clinical decision-making. They pointed out that these models can hallucinate facts or miss critical clinical context. If an AI agent drafts an incorrect medication order, the physician is still the one signing it and taking the legal responsibility.
Marques M et al, *Cureus* 2024, explored the ethical challenges of machine learning algorithms in decision-making. The medicine revolution through artificial intelligence brings up serious questions of liability and bias. If an algorithm is trained on skewed demographic data, the clinical AI agents will propose biased clinical actions that could harm specific patient populations.
The physician as the final checkpoint
You can never completely remove the human element from medicine. The AI is a tool, not a doctor. It is essential to seek proper verification for every AI-generated suggestion before applying it to patient care. Your medical license is the one on the line when things go wrong. The Penn Medicine deployment specifically keeps the physician in the loop for review and approval for this exact reason. The agents propose clinical actions, but they do not execute them autonomously. You have to watch out for false positives and generic advice that does not fit the specific patient in front of you.
Conclusion
Undoubtedly, the shift from ambient scribes to clinical AI agents is going to alter the daily practice of medicine forever. If you have suffered from extreme documentation fatigue, this technology offers a real path forward. The partnership between Penn Medicine and K Health proves that major academic institutions are ready to trust active AI workflows with real patient data. The research shows that these systems reduce charting time and help manage complex clinical data across multiple specialties. While there is no specific cure for the administrative bloat in modern healthcare, clinical AI agents can certainly help manage the condition properly. If your health system is considering this leap, you can rest assured that the tools are becoming much more capable of handling the heavy lifting.
References
- Huang K et al. Biomni: A General-Purpose Biomedical AI Agent. bioRxiv : the preprint server for biology 2025. doi:10.1101/2025.05.30.656746 (PMID: 40501924)
- Marques M et al. The Medicine Revolution Through Artificial Intelligence: Ethical Challenges of Machine Learning Algorithms in Decision-Making. Cureus 2024. doi:10.7759/cureus.69405 (PMID: 39411643)
- Truhn D et al. Artificial intelligence agents in cancer research and oncology. Nature reviews. Cancer 2026. doi:10.1038/s41568-025-00900-0 (PMID: 41526721)
- Hager P et al. Evaluation and mitigation of the limitations of large language models in clinical decision-making. Nature medicine 2024. doi:10.1038/s41591-024-03097-1 (PMID: 38965432)
- Guo Y et al. From Images to Genes: Radiogenomics Based on Artificial Intelligence to Achieve Non-Invasive Precision Medicine in Cancer Patients. Advanced science (Weinheim, Baden-Wurttemberg, Germany) 2025. doi:10.1002/advs.202408069 (PMID: 39535476)
- Liu S et al. Self-driving bioprinting laboratories. Biofabrication 2026. doi:10.1088/1758-5090/ae3645 (PMID: 41512329)
- Goldstein J et al. Determinants for scalable adoption of autonomous AI in the detection of diabetic eye disease in diverse practice types: key best practices learned through collection of real-world data. Frontiers in digital health 2023. doi:10.3389/fdgth.2023.1004130 (PMID: 37274764)
- Román-Belmonte JM et al. Artificial intelligence in musculoskeletal conditions. Frontiers in bioscience (Landmark edition) 2021. doi:10.52586/5027 (PMID: 34856771)
- Kazemzadeh K. Artificial intelligence in ophthalmology: opportunities, challenges, and ethical considerations. Medical hypothesis, discovery & innovation ophthalmology journal 2025. doi:10.51329/mehdiophthal1517 (PMID: 40453785)
- Khosravi B et al. Agentic AI in Radiology: Evolution from Large Language Models to Future Clinical Integration. Radiology. Artificial intelligence 2026. doi:10.1148/ryai.250651 (PMID: 41532836)
- Shah SJ et al. Ambient artificial intelligence scribes: physician burnout and perspectives on usability and documentation burden. Journal of the American Medical Informatics Association : JAMIA 2025. doi:10.1093/jamia/ocae295 (PMID: 39657021)
- Ma SP et al. Ambient artificial intelligence scribes: utilization and impact on documentation time. Journal of the American Medical Informatics Association : JAMIA 2025. doi:10.1093/jamia/ocae304 (PMID: 39688515)
- Hassan H et al. Clinical Implementation of Artificial Intelligence Scribes in Health Care: A Systematic Review. Applied clinical informatics 2025. doi:10.1055/a-2597-2017 (PMID: 40306686)
- Duggan MJ et al. Clinician Experiences With Ambient Scribe Technology to Assist With Documentation Burden and Efficiency. JAMA network open 2025. doi:10.1001/jamanetworkopen.2024.60637 (PMID: 39969880)
- Shah SJ et al. Physician Perspectives on Ambient AI Scribes. JAMA network open 2025. doi:10.1001/jamanetworkopen.2025.1904 (PMID: 40126477)
- https://www.fiercehealthcare.com/ai-and-machine-learning/penn-medicine-k-health-partner-deploy-clinical-ai-clinical-agents
Licensed physician and clinical AI specialist. Founder and Editor-in-Chief of ZayedMD, a physician-led medical publication covering clinical AI, neurology, metabolic health, and evidence-based patient guidance.
