AI Scribes: From Documentation Tool to Clinical Copilot
AI medical scribes are quickly establishing themselves as essential tools in U.S. healthcare. Consulting firm McKinsey estimates that over 10% of U.S. physicians already use AI scribe solutions, with companies such as Abridge, Ambience, Suki, Nabla, and Microsoft among the leading technology vendors. Major electronic health record (EHR) players including Epic, Athenahealth, and Oracle have also begun integrating AI scribe capabilities directly into their platforms.
According to a feature by Healthcare Brew, Jon Ebbert, MD, who oversees voice programs at Mayo Clinic in Rochester, Minnesota, believes AI scribes will “transform care.” (Healthcare Brew article: https://www.healthcare-brew.com/stories/what-are-ai-scribes) Market research firm Grand View Research forecasts that the AI scribe market will expand from 397.1 million dollars in 2024 to roughly 3 billion dollars by 2033, reflecting the speed at which this technology is diffusing across health systems.
What AI Scribes Actually Do
For every encounter, physicians must record the patient’s condition and treatment plan and upload that documentation to the EHR. This requirement has long been cited as a major driver of burnout. Physicians often continue documentation at home after hours, a phenomenon widely known as “pajama time.” A 2016 study by the American Medical Association (AMA) found that doctors were spending more time on EHR and desk work than with patients. In a 2024 AMA survey update of 18,000 physicians, 43.2% reported at least one symptom of burnout.
AI scribes are designed to directly target this problem. Matt Troup, Senior Director of Clinical Success at Abridge, describes the technology as an “intelligence system for clinical conversations.” Abridge’s solution has been deployed at more than 200 institutions, including Mayo Clinic, Kaiser Permanente, Memorial Sloan Kettering Cancer Center, and UChicago Medicine.
Functionally, the concept is relatively simple. During a consultation, the AI scribe listens to the real‑time conversation between clinician and patient. It analyzes what questions the physician asks, what diagnoses are made, and what treatment plan is discussed. From there, it generates a transcript and uses that transcript to draft a clinical note suitable for entry into the EHR. Final accountability, however, remains squarely with human clinicians. Physicians must review and edit the draft before confirming it in the record to ensure clinical accuracy and satisfy legal and regulatory requirements.
Saving Time and Restoring “Eyes on the Patient”
The most intuitive benefit of AI scribes is time savings. A 10‑week pilot study highlighted in the New England Journal of Medicine, involving 3,442 physicians and 300,000 encounters, found that AI scribes reduced total documentation time by 15,700 hours over the trial period. Jon Ebbert emphasizes that this time reduction directly translates into more “eyes on the patient.”
He notes that the tool reduces concern about documentation during the visit itself. In his own practice, using AI scribe tools has allowed him to keep his gaze on the patient longer and concentrate more fully on the conversation, rather than splitting attention between the patient and the computer screen.
What Clinicians Want from AI Scribes
Mayo Clinic employs roughly 83,000 staff across 16 hospitals, 45 multidisciplinary clinics, and a mobile clinic. Over the past year and a half, it has piloted Abridge and Ambience across its network. Currently, about 4,200 clinicians are using AI scribe tools, and the institution is actively working to bring more clinicians on board.
For Ebbert, one of the most important selection criteria is customizability of the output. The documentation style and note structure needed by a primary care physician differ substantially from those required by a sub‑specialist. “Everyone needs something different, and everyone wants their notes to look a certain way,” he explains.
Another key criterion is avoiding over‑aggressive summarization. Some tools compress the clinician’s reasoning into a few sentences or bullet points, which can strip away the context of clinical judgment. Ebbert says physicians want AI scribes to richly reflect their cognitive processes without reducing that complexity to a handful of overly simplified lines. Preserving the nuance of medical reasoning matters not just for care quality but also for medico‑legal clarity.
Beyond “Notes”: Toward a Full Clinical Copilot
AI scribe vendors are already moving beyond simple documentation assistance and positioning their tools as copilots for the entire care journey. “To be honest, what we do now goes far beyond notes,” says Abridge’s Troup. While Abridge initially focused on note generation, the company now sees broader opportunities to apply AI to administrative workflows before and after the encounter.
One example is Ambience, which launched a copilot product in August 2025. The solution surfaces the patient’s historical record in real time during the visit so that clinicians can make decisions with immediate contextual awareness of prior diagnoses, test results, and treatments. Looking ahead, Ebbert expects much deeper integration with EHR systems. In his vision, when a clinician orders a blood test during the consultation, an ambient AI system will detect the conversation and automatically place the order into the workflow.
He also highlights AI‑powered translation as a major opportunity. If patients and clinicians who speak different languages could converse without a language barrier, the quality and accessibility of care would improve substantially, especially in diverse or multilingual regions.
After three decades in practice, Ebbert calls this “the biggest change” to his day‑to‑day care delivery. In his words, no prior technology has helped him end his day “on time and in balance” as effectively as AI scribe tools. By returning documentation time to direct patient care and clinical reasoning—and by automating and standardizing more of the care journey through tight EHR integration—AI scribes are emerging not just as note‑taking tools but as infrastructure reshaping how modern healthcare operates.
The Rise of AI Scribes and Voice Documentation in Korea
AI‑driven, voice‑based medical documentation is also gaining momentum in South Korea. Asan Medical Center in Seoul announced that it has built the country’s first AI‑based clinical voice recognition system that records and summarizes conversations between clinicians and patients in real time and automatically saves them as medical records across emergency rooms, wards, and outpatient clinics. The system goes beyond simple speech‑to‑text transcription, performing real‑time transcription, symptom extraction, disease classification, and conversation summarization, all integrated into the hospital’s AMIS 3.0 medical information system so that structured notes can be automatically stored in the EMR. (Asan Medical Center news release: https://news.amc.seoul.kr/news/con/detail.do?cntId=10524, English version:news-en.amc.seoul]
Seoul St. Mary’s Hospital has begun piloting a next‑generation AI documentation solution called “CMC GenNote,” developed with healthcare startup PuzzleAI. The system uses a large language model optimized for clinical environments and supports opening documentation templates by voice command, then transforming spoken content into structured EMR entries mapped to each template field. The goal is to reduce administrative burden on trainees and attending physicians alike compared with earlier “Voice EMR” systems that focused mainly on radiology or report transcription. (Example coverag, digitalhealthnews]
Korean AI healthcare company Tesser has launched “Ontol Scribe,” a solution that automatically analyzes health check‑up results and generates comprehensive summary opinions. The product is being rolled out to local hospitals such as Seoul Yangji Hospital and Incheon Ain Hospital, with the company targeting adoption in approximately 50 hospitals by the end of the year. (Product and deployment details: https://platum.kr/archives/250531) By automating report generation, Ontol Scribe significantly reduces the workload of staff in health screening centers and illustrates how the AI scribe concept can extend beyond point‑of‑care visits into preventive care, billing, and other administrative domains.youngiverse.tistory+1
Another prominent voice‑focused example is Selvas AI’s “MediVoice,” which supports real‑time voice‑driven medical imaging reports and clinical documentation in radiology, operating rooms, and gastroenterology departments.
The company reports that MediVoice enables report entry more than three times faster than manual transcription and achieves speech recognition accuracy above 98%. Looking forward, Selvas AI aims to integrate functions such as collaborative scheduling, insurance claims, and discharge instructions into a single intelligent agent architecture that orchestrates multiple administrative workflows within hospitals. (Overview of MediVoice and roadmap:
Taken together, these U.S. and Korean developments—Abridge, Ambience, Suki, and Microsoft Nuance DAX in the United States, and deployments at Asan Medical Center, Seoul St. Mary’s Hospital, and multiple domestic AI startups in Korea—demonstrate a converging global trend. Rather than “replacing doctors,” AI is increasingly taking over documentation and administrative burdens, acting as a copilot that returns time to clinicians and supports more focused, relational, and data‑informed patient care.news-en.amc+3