The AI medical scribe glossary

When an AI model writes something confidently into a note that was never said.

Software that drafts clinical notes from a conversation, replacing or assisting a human scribe.

AI that listens to a clinical encounter and drafts a structured note automatically.

The speech-to-text foundation under every ambient scribe.

The transformer-based language model under every AI scribe.

AI that runs entirely on the clinician's own machine, so data never reaches a cloud.

Pulling supporting documents into the AI's context before it answers.

Separating who said what in a multi-speaker recording.

Therapy note format: Behaviour, Intervention, Response, Plan.

Therapy note format: Data, Assessment, Plan.

Therapy note format: Goal, Intervention, Response, Plan.

Structured account of the patient's current symptoms — the heart of the Subjective section.

Structured psychiatric assessment of appearance, behaviour, mood, thought, cognition.

Therapy note format: Problem, Intervention, Response, Plan.

Structured running list of a patient's active and resolved diagnoses in the EHR.

Standard four-part clinical note: Subjective, Objective, Assessment, Plan.

Contract that makes a vendor legally responsible for HIPAA-protected data they touch.

Where the vendor actually stores and processes your data.

GDPR-required contract between you (controller) and the vendor (processor).

Switzerland's revised Federal Act on Data Protection — stricter than GDPR in some respects.

EU data-protection law — non-negotiable for any EU healthcare deployment.

US federal law protecting health information; the minimum compliance bar for US deployments.

Healthcare-specific control framework combining HIPAA, NIST and ISO requirements.

International standard for information-security management systems.

Protected (electronic) Health Information — the data HIPAA covers.

Independent audit of a vendor's security controls over a period of time.

A third party the vendor uses to process your data (cloud, LLM provider, etc.).

Whether the vendor uses your sessions to train its AI models.

A policy where vendor immediately deletes audio and / or notes after use.

The MDR rule that classifies decision-informing healthcare software as a regulated device.

Mark indicating an EU medical device has passed conformity assessment.

Lowest MDR risk class — self-declared, no Notified Body required in most cases.

Moderate-risk MDR class — Notified Body assessment and CE mark required.

EU regulation classifying AI by risk; high-risk AI in regulated products faces extra obligations.

EU Medical Device Regulation — governs healthcare software that has a medical purpose.

US carve-out exempting transparent clinical decision support from FDA device regulation.

European Commission guidance on qualifying and classifying medical-device software under MDR.

Independent EU-designated organisation that audits regulated medical-device software.

American Medical Association codes for procedures and services billed in US healthcare.

Evaluation-and-management billing — the CPT subset most outpatient visits use.

International Classification of Diseases (10th revision) — standard diagnosis codes.

Newer WHO diagnosis-code standard, gradually being adopted globally.

US clinician identifier — used by some tools to gate access to verified clinicians.

Standardised clinical terminology — more granular than ICD, used in EHRs.

Swiss outpatient tariff replacing TARMED — relevant for Swiss-built scribes.

Per-statement provenance linking note content to the moment it was generated from.

Scribe-generated content flowing back into the EHR's structured fields.

A scribe running inside the EHR's own UI, not as a separate app.

AI-generated post-visit summary intended for the patient.