The promise around ambient AI scribes is straightforward: let the technology capture and structure the consultation, and give clinicians more attention for the patient. A large real-world study from a Spanish healthcare network suggests that the implementation question is more interesting than the promise. AI-assisted documentation can be adopted at scale and improve elements of the clinical record, but that does not automatically translate into shorter consultations.
The study, published in Scientific Reports, examined more than 2.3 million outpatient consultations across 45 specialties between October 2024 and December 2025. It assessed an AI scribe used in routine care within the Quironsalud network in Spain. The scale makes the research notable. Much of the evidence around ambient documentation still comes from pilots, small departmental deployments or self-reported experience. Here, the authors could examine adoption, record quality and workflow patterns across a large and varied clinical setting.
A record is part of care, not just administration
Clinical documentation supports continuity, safety, coding and the patient conversation after the consultation. When the record is incomplete, ambiguous or finished late, the consequences reach beyond administrative work. The study reports stronger documentation in a manual audit of past medical history and allergy status in scribe-supported records. It also examined semantic agreement, readability and structured quality.
Those findings are relevant because they shift the discussion from a familiar but narrow metric: minutes saved. An AI scribe may add value when it helps a clinician produce a more complete and usable record, even if the consultation itself is not shorter. Conversely, a tool that creates a note quickly but introduces omissions, unclear language or a new verification burden does not improve care simply because it is called efficient.
The reported consultation duration was not lower in the assisted group. That does not prove that the technology makes care slower; the study is observational and consultation types, specialties and adoption patterns differ. It does, however, challenge the assumption that an ambient scribe will deliver the same productivity gain in every setting. The authors also work within the healthcare network studied, which makes independent replication and continued evaluation important.
Scaling changes the question
The strongest signal in the Spanish experience is that adoption rose from a small share of outpatient activity to around one third of consultations over sixteen months. That is no longer an experiment at the edge of a health system. It is a change to clinical work that needs training, workflow design, data protection, quality assurance and clear professional accountability.
At that scale, implementation choices matter as much as the model itself. Professionals need to know when they remain responsible for reviewing a generated note, how corrections are handled, where data are processed and whether the tool works equally well across specialties, languages and patient populations. Organisations need to measure the right outcomes: not only usage, but documentation quality, patient experience, time outside normal hours, safety incidents and impact on the work of clinical teams.
From AI feature to professional infrastructure
Ambient documentation is often presented as a quick route out of administrative burden. The Spanish study points to a more realistic conclusion. AI can support professionals, but only when it becomes part of a well-designed clinical process. The aim should not be to automate the consultation. It should be to protect the professional attention that makes a consultation valuable, while making the resulting record safer and more useful.
That is also the more important question for health-system leaders. The next phase of healthcare AI is not about demonstrating that a model can generate text. It is about establishing which tools improve work and care in routine practice, under what conditions and with what safeguards. Large-scale evidence from Spain adds a useful data point, but it is not a universal answer. It shows why implementation has to be evaluated as rigorously as the technology.
At the annual ICT&health World Conference 2027, AI is approached as a clinical, governance and workforce question across three days: from application in practice to responsible scale-up and the impact on the professional workplace.
Add ICT&health on Google
Show more content from ICT&health in Google Search.
