Clinical Safety Boundaries and Production Readiness

Understand AfriHealth AI’s human-in-the-loop safeguards, what the prototype does not do, and the readiness gates required before using real patient data.

AfriHealth AI is built to make clinical speech easier to document, never harder to verify. Every transcript and generated artifact is a draft for a qualified clinician. This page explains the safety boundaries of the prototype and the gates that remain open before operational use.

Human-in-the-loop by design

• Transcripts, triage summaries, SOAP drafts, and ICD-10 suggestions are drafts until a clinician verifies them.
• Module 2 requires Verify & Sign Off before any EHR commit.
• Triage and follow-up support decisions. They do not diagnose, contact patients autonomously, or dispatch care.

Current status

Area	Status
Prototype	Functional
Production	Not ready for identifiable data
Provider credentials	Kept server-side
Private validation inputs	Kept outside Git

Open readiness gates

Do not use the prototype with identifiable patient data until these gates are closed.

• Transcript retention and deletion controls.
• Built-in clinician authentication and role-based access.
• Verified production gateway configuration and exact production CORS origins.
• End-to-end medication safety enforcement. Until then, verify all medication details independently.

Delivery priorities

1. Close production safety gates: retention/deletion, authentication, and gateway checks.
2. Complete clinician-reviewed workflows: validate transcript-to-SOAP and coding, and add a backend medication safety gate.
3. Strengthen evaluation evidence: expand consented, de-identified Amharic-English evaluation.
4. Verify release and handoff: confirm deployment and complete the team handoff checklist.

Related pages

• Benchmarks
• Voice EHR Intake