Clinical documentation & AI scribe glossary
Definitions of the terms that come up when a clinic evaluates ambient AI documentation — written for clinicians and practice managers, not engineers.
- Ambient AI scribe
- Software that listens to a consultation in the background, transcribes it, and drafts a structured clinical note without the clinician dictating to it. The clinician talks to the patient normally; the scribe works from that conversation and produces a draft for review and signature.
- SOAP note
- The standard structure for a clinical encounter note: Subjective (what the patient reports), Objective (examination findings and measurements), Assessment (the clinician's impression or differential) and Plan (investigations, treatment and follow-up). Most AI scribes output this format by default.
- ASR (automatic speech recognition)
- The engine that converts audio into text. Clinical ASR is judged on how well it handles medical vocabulary, drug names, accents and background noise — general-purpose ASR usually degrades on all four in a real consultation room.
- Speaker diarisation
- Separating a recording into who said what. Diarisation matters clinically because a symptom reported by the patient and a finding stated by the clinician belong in different sections of the note.
- Zero-footprint audio
- An architecture where raw consultation audio is transcribed and then discarded rather than written to long-term storage or used to train models. It reduces the surface area of a breach: there is no archive of patient voices to leak.
- Code-switching
- Alternating between two or more languages within a single conversation or sentence — common in multilingual clinics. A scribe that handles code-switching can follow a consultation that moves between, say, Yorùbá and English and still produce one coherent note in clinical English.
- Clinical documentation burden
- The share of a clinician's working time spent writing, structuring and filing notes rather than treating patients. It is the main driver of after-hours charting and a recognised contributor to burnout, and it is the cost an ambient scribe is meant to reduce.
- Pyjama time
- Informal term for documentation completed at home after clinic hours. It is a practical proxy for documentation burden: if a tool works, pyjama time falls.
- EMR / EHR
- Electronic Medical Record and Electronic Health Record — the system of record for patient charts. EMR usually refers to records within one practice; EHR implies records shared across organisations. A scribe integrates with these either by copy-and-paste, structured export, or a mapped field integration.
- HL7 FHIR
- A modern healthcare data exchange standard defining resources such as Patient, Encounter and DocumentReference. FHIR support is what allows a note generated outside the EHR to be filed inside it as structured data rather than as a flat block of text.
- ICD-10 / ICD-11
- International Classification of Diseases coding systems used to record diagnoses for statistics, billing and reimbursement. AI coding assistants suggest candidate codes from the note; the clinician or coder confirms them.
- CPT code
- Current Procedural Terminology codes describing the services and procedures performed in an encounter, used mainly for billing in the United States. Suggested codes are draft content and must be verified against what was actually done.
- Medical necessity / justification letter
- A written argument that a requested test, medication or procedure is clinically warranted, usually sent to a payer or scheme. AI can draft it from the encounter record, but the clinician owns the clinical claim it makes.
- PHI (protected health information)
- Any health information that can be tied to an identifiable individual, including audio of a consultation. How a vendor stores, transmits and deletes PHI is the core of any compliance assessment.
- BAA (business associate agreement)
- A contract under HIPAA in which a vendor handling PHI on a covered entity's behalf accepts defined safeguard and breach-notification obligations. Without one, a US clinic cannot lawfully route PHI through the vendor.
- Human-in-the-loop
- A design principle in which AI output is never applied automatically: a qualified person reviews, corrects and approves it. For clinical documentation it means the signed note is the clinician's, not the model's.
- Hallucination
- A model producing content that was never said or that does not exist — an invented symptom, dose or finding. It is the central safety risk in AI documentation and the reason every draft note must be read against the encounter before signing.
- Note quality score
- A measure of how complete and internally consistent a note is — whether each SOAP section is populated, whether the plan follows the assessment, and whether anything contradicts itself. Useful as a review prompt, not as a substitute for reading the note.