Modern technologies should not dehumanize patient care but instead help doctors reconnect with patients by allowing more face-to-face interaction without screens getting in the way. The idea appears repeatedly in digital health discussions and reflects a core principle: AI should help reduce one of physicians’ biggest burdens, excessive bureaucracy. Clinical copilots may offer a part of this solution.

A 2017 study from the University Hospital of Lausanne, published in the Annals of Internal Medicine, estimated that physicians spent an average of 1.7 hours/shift with patients, 5.2 hours using computers, and 13 minutes doing both simultaneously. Researchers have also found that doctors spend approximately 52% of their time on tasks indirectly related to patient care, including medical documentation.

Nearly a decade later, the emergence of clinical copilots may have begun to change this pattern. Ambient transcription systems combined with generative AI can now convert patient-clinician conversations directly into structured medical documentation in real time, including clinical notes, referral letters, and medical records.

Speaking with El Médico Interactivo, César Dilú Sorzano, MD, a member of the Digital Health Group of the Spanish Society of General and Family Physicians. “There is already fairly consistent evidence showing that these systems can reduce a significant share of the workload currently placed on doctors and nurses, particularly administrative tasks.”

Reducing Burden

According to a 2025 study from Stanford University, ambient AI transcription systems based on advanced language models reduced documentation burden, physician burnout, and improved usability.

“In a subsequent multicenter study involving 263 professionals across six healthcare systems, the proportion of clinicians experiencing burnout dropped from 51.9% to 38.8% after 30 days of use,” Sorzano said. “Put simply, the copilot does not remove patients or reduce clinical complexity, but it can reduce part of the repetitive work that currently consumes clinicians’ time and energy.”

During the third Spanish-Portuguese Primary Care Physicians Conference organized by the Extremadura Society of General Practitioners and Family Physicians, in collaboration with the Portuguese Medical Association, Sorzano said that “Generative AI can save between 1 and 2 hours of administrative work per day, allowing the family doctor to dedicate more time to what really matters: direct patient care.”

“It should not be presented as a magic solution because the benefits depend heavily on how it is implemented and whether it truly saves time overall or simply adds another layer of review,” he said.

2026 Outlook

Additional findings came from a 2025 study sponsored by National Health Service London, England, which showed that the use of clinical copilots was associated with a 23.5% increase in direct interaction time with patients during appointments, along with an 8.2% reduction in total appointment duration when AI-powered medical assistants were used. Emergency departments showed particularly strong results, with a 13.4% increase in the number of patients seen per shift.

“When implemented properly, AI can improve the experience on both sides of the consultation,” Sorzano added. “But there is another possible interpretation. In some settings, efficiency gains may be used to increase patient volume rather than to improve quality. In fact, the departments in this London study saw more patients per shift. This means that the debate is not only technological, but also ethical and organizational. The key question is not whether AI frees up time but who decides how that time is used.”

Another important question is whether these clinical copilots not only save time but also improve the quality of care. Sorzano explained that these systems “can perform several useful and highly specific tasks: listening to the consultation, transcribing it, summarizing the patient’s history, organizing information, generating a clinical note, preparing letters or reports, and, in some cases, identifying possible omissions or medication alerts for the clinician to review. What these systems currently do best is not ‘thinking for the doctor’ but reducing time spent on documentation.”

However, Sorzano cautioned that these tools could still make mistakes. They may omit information, misinterpret statements, or even include details that have never been mentioned. In other words, they can support clinical practice but are not substitutes for clinical judgment.

With burnout continuing to rise among healthcare professionals, clinical copilots may also help reduce not only physical workload but also mental strain.

Much of the available research has focused on the mental burden associated with clinical documentation. “The Stanford study used a validated scale, the NASA Task Load Index, and the reduction was clear,” Sorzano said.

According to Sorzano, similar studies also showed improvements in cognitive workload linked to clinical notetaking, patient-centered attention, and time spent documenting outside working hours. “Free text tools help because they allow professionals to speak and think naturally while the system transforms the conversation into an organized draft,” he explained. “What they really reduce is multitasking, and that is truly exhausting.”

Sorzano emphasized that these systems still have important limitations. “Reducing the mental burden linked to documentation does not, by itself, solve the healthcare system’s underlying problems,” Sorzano said. “If there is a shortage of staff, excessive patient volume, or overloaded schedules, AI does not resolve those structural issues.”

Key Elements

Another challenge is the wide variety of tools currently available. There is no single model that makes it difficult to evaluate these systems as a uniform group. When asked what defines a truly effective model, Sorzano outlined several essential features.

“It has to be a tool that healthcare professionals can understand and control”, Sorzano said. “At a minimum, that means five things: seamless integration with the medical record, the ability to easily correct generated content, a clear record of what the system has done, transparency about what it is and is not designed to do, and robust safeguards for patient privacy and data security.”

According to Sorzano, these elements are essential for earning the trust of healthcare professionals. “If the system appears to be an opaque black box that generates content that is difficult to audit, mistrust is a natural response,” he said. “In Spain, that trust also depends on public governance.” He noted that the Digital Strategy for AI in the Spanish National Health System already outlines a framework focused on safe, useful, and interoperable implementation supported by professional training.

Adequate training is another critical factor that Sorzano added. “Without training, even a good tool can fail.”

Sorzano concluded that Spain should avoid both excessive caution and reckless adoption. “The sensible approach is careful implementation: rigorously evaluated pilot programs, strong data protection, integration with electronic health records, professional training, and mandatory human oversight”, he said. “That approach aligns most closely with the current European framework and with the direction already proposed in the Spanish National Health System’s strategy.”

Sorzano reported having no relevant conflicts of interest.

This article was translated from El Médico Interactivo on Univadis, part of the Medscape Professional Network.