Glossary · term

Ambient scribe

An ambient scribe is a clinical documentation tool that captures a clinician–patient conversation with little interaction, converts speech to a transcript, and typically uses generative AI to produce a structured draft note or letter. `Ambient` describes capture during the encounter; it does not mean constant or undisclosed recording. The draft is not the final medical record: a responsible clinician reviews, edits and authorizes what is retained.

Other2023-09Wave 2 · 2024Maturity: 4/5

Origin and context

Automated dictation and speech recognition predate generative AI. The earliest exact `Ambient Scribe` wording located in this review appears as a Tali product-feature name in a September 2023 Canadian health-technology issue; that is an evidence anchor, not a coinage claim. A March 2024 NEJM Catalyst report then used `ambient AI scribes` for a large Kaiser Permanente deployment. By 2025–2026, NHS England and Australia's national safety commission used the label generically in official guidance.

Sources: s1, s2, s3, s7

Why it matters

The workflow shifts documentation from typing or dictating a note after the visit to reviewing a machine-generated draft. This can change where effort occurs and how much attention a clinician gives the screen. Evidence does not support a universal efficiency claim: a 2025 systematic review found small, heterogeneous studies and inconsistent system-level results, while a 2026 prospective Dutch study measured less documentation time but no change in total consultation time.

Sources: s4, s5, s6

Example

A clinician starts capture for an encounter, speaks with the patient normally, then receives a transcript-derived draft organized into the local note template. The clinician checks names, medications, symptoms, examination findings, diagnoses and plans, removes irrelevant third-party details, corrects omissions or invented text, and only then signs or transfers the note to the health record. Recording, storage and EHR integration vary by product and setting.

Sources: s3, s5, s7

How it differs

Ambient Clinical Intelligence (ACI)

`Ambient Clinical Intelligence (ACI)` was introduced by Nuance in 2019 as a broader vendor-framed category spanning documentation, assisted workflows, task and knowledge automation, and clinical guidance. Ambient scribing is the narrower, vendor-neutral documentation workflow. The terms overlap but are not exact aliases, so the existing ACI record should remain a related reference rather than be redirected or merged automatically.

AI hallucination

`Hallucination` is one possible output failure—fabricated or nonsensical content—not another name for the system. Ambient scribes can also omit, mishear, over-summarize or over-expand information, so checking only for fabricated facts is insufficient.

Maturity and evidence

The term merits maturity 4 for category adoption, not for proven clinical effectiveness. It appears in official guidance from independent health authorities and in studies of deployments across multiple health systems and countries. The underlying workflow is established enough to define consistently, while product performance, outcome measures and governance practices remain uneven.

Sources: s2, s3, s4, s5, s6, s7

Limits and open questions

A category label does not establish that a product is safe, accurate, compliant, integrated with an EHR, or regulated in a particular way. Studies report editing burden, verbosity, missing or incorrect details, multilingual and accessibility problems, patient discomfort, privacy concerns and possible interference with clinical reasoning. Consent, transparency, data retention and device-regulation requirements depend on jurisdiction, intended use and local policy. This page explains the workflow; it does not advise clinicians or organizations how to deploy it.

Sources: s3, s4, s5, s6, s7

Related terms

References

Last updated: 2026-09-07

In the Skills Atlas

This term is also covered in the Skills Atlas as human in the loop ai skill.

In the Skills Atlas

This term is also covered in the Skills Atlas as ai risk management skill.

In the Skills Atlas

This term is also covered in the Skills Atlas as ai guardrails skill.