AI safety & governance
AI drafts. People decide.
Every AI output in Narzim passes through a person before it reaches the chart or the patient. Owners set the limits; the activity log shows every step.
Authority classes
Each AI tool has a fixed job and a fixed limit.
A tool can never act above its class. Select a class to see what it can and cannot do.
ScribeScribes turn a consented conversation into a draft. The draft stays editable and unsigned until the doctor approves it.
CanDraft SOAP notesSuggest diagnosis codesDraft prescriptions
CannotRecord without consentSign or save a noteIssue a prescription
Owner controls
Your clinic sets the rules. Try them.
These are the settings clinic owners see in Settings → AI. Every change is recorded with who made it and when.
Require consent before recordingScribe cannot start until consent is captured.
Review before savingAI drafts stay unsigned until a clinician approves.
Advanced model tierUse the larger model for complex specialties.
Keep consultation audio
Our commitments
What we promise about AI.
Always labelledAnything AI wrote is marked until a person approves it.
Consent recordedPatient consent to recording is stored with the visit.
Separate by designNarzim GPT never has access to patient records.
Fully traceableEvery AI action appears in the chart’s activity timeline.
Consent
The patient says yes before anything is recorded.
Scribe shows a clear request on screen. The answer, the time and who gave it are saved with the visit. If the patient says no, the doctor writes the note as usual.
AI governance questions
Narzim uses leading large language models through enterprise agreements. Owners choose the standard or advanced tier in Settings.