A Technology on the Rise
Ambient AI is beginning to reshape everyday clinical care, most visibly through the rapid uptake of AI “scribes” that listen to patient–clinician conversations and generate draft clinical notes. These tools use speech recognition and large language models to record and summarise the encounter, with some health systems already deploying them at scale to help reduce documentation workload and free clinicians to focus more fully on patients.
The Importance of a Balanced View
As this technology becomes more common, it raises important questions alongside its promise. Research highlights potential benefits, such as easing administrative burden and improving the quality and consistency of documentation. However, concerns also emerge, including the loss of “human” nuance in clinical notes, ethical challenges around data handling, and the possibility that these systems might also enable new forms of clinician surveillance. Recent studies discuss how Ambient AI might support clinical work, where caution is warranted, and what responsible adoption should look like.
Potential Benefits
Proponents of Ambient AI argue that these tools offer meaningful practical benefits in busy clinical settings. Research to date points to several potential advantages for AI scribes:
- Reduced documentation workload
- Less after-hours charting
- More accurate and consistent notes
- Greater attention to the patient during the clinical encounter
Beyond the draft notes they produce, AI scribes also generate full transcripts of clinical encounters, which (if retained) could support quality‑improvement efforts, research on clinical communication, and a clearer understanding of how patients describe their symptoms and concerns. Several scholars have also noted that ambient systems may eventually help strengthen patient‑centred care by capturing more of the patient’s authentic voice, provided the technology is designed and governed in ways that value and preserve those perspectives.
Concerns for Patient Care
Alongside their potential benefits, Ambient AI systems raise important concerns about the quality of patient care. Some AI scribes are designed to filter out what they classify as non‑essential conversation, but clinicians note that these seemingly informal exchanges often contain crucial personal details that help them understand a patient’s circumstances and build trust over time. At the same time, many health systems are choosing to automatically delete AI‑generated transcripts because of malpractice worries, limiting transparency and making it difficult to evaluate how accurately these tools capture clinical encounters. Studies also highlight patient‑facing challenges, including varied levels of comfort with passive audio recording, concerns about sensitive information being captured, and uncertainty around how informed consent should be obtained in real‑world practice.
More robust patient-facing materials, co-designed with diverse patients, might have better illuminated how ambient systems capture, process, and store conversational data, empowering patients to make informed choices about participation (Griffen et al., 2026)
Concerns about Clinician Surveillance
Ambient AI can also create new forms of oversight that extend far beyond documentation support. Because these systems analyse speech patterns, timing, and the structure of clinical interactions, they can generate detailed data about how clinicians communicate, how closely they follow institutional protocols, and even how much time they spend with each patient. Researchers warn that such capabilities could easily shift from quality‑improvement tools to mechanisms of surveillance, contributing to the emergence of “quantified workers” in medicine and reducing clinicians’ professional autonomy.
“quantified workers” — workers whose daily tasks are monitored and controlled by AI technologies (Cohen et al., 2025)
These risks may be felt most acutely by those with less institutional power, raising broader workplace‑ethics concerns about fairness, trust, and the conditions under which Ambient AI is introduced into clinical practice.
Towards Responsible Implementation
Taken together, the research suggests that Ambient AI holds meaningful potential for improving clinical work, but only if its adoption is guided by careful design and oversight. Ensuring responsible use will require transparency about how these systems operate, strong governance structures, thoughtful data‑stewardship practices, and safeguards that prevent misuse, whether in patient care or in the management of clinicians. These goals align closely with the mission of AI2MED, which aims to examine AI in clinical settings with evidence, nuance, and ethical attention. By approaching Ambient AI with both optimism and caution, health systems can work toward maximising its benefits while protecting the people who rely on it: clinicians, patients, and the relationships at the centre of care.
Bowker, D., Torti, J., & Goldszmidt, M. (2023). Documentation as composing: how medical students and residents use writing to think and learn. Adv Health Sci Educ Theory Pract, 28(2), 453–475. https://doi.org/10.1007/s10459-022-10167-x
Cohen, I. G., Ajunwa, I., & Parikh, R. B. (2025). Medical AI and Clinician Surveillance — The Risk of Becoming Quantified Workers. New England Journal of Medicine, 392(23), 2289–2291. https://doi.org/doi:10.1056/NEJMp2502448
Goodman, K. E., & Morgan, D. J. (2026). Digital Exhaust or Digital Gold? The Value of AI-Generated Clinical Visit Transcripts. New England Journal of Medicine, 394(2), 110–113. https://doi.org/doi:10.1056/NEJMp2514616
Griffen, Z., Lawrence, K., & Owens, K. (2026). Justice Begins in the Field: How Empirical Data Can Inform Ethical Analysis of Ambient Intelligence Systems. The American Journal of Bioethics, 26(2), 29–32. https://doi.org/10.1080/15265161.2025.2608634
Leung, T. I., Coristine, A. J., & Benis, A. (2025). AI scribes in health care: balancing transformative potential with responsible integration. JMIR Medical Informatics, 13(1), e80898.
Schiff, G. D. (2025). AI-Driven Clinical Documentation — Driving Out the Chitchat? New England Journal of Medicine, 392(19), 1877–1879. https://doi.org/doi:10.1056/NEJMp2416064
Image credit: Fanny Maurel & Digit / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0/
