Governance Engine

Every answer passes through five safety gates.

Dana doesn't just generate text — it routes every patient question through a structured pipeline that classifies intent, selects a guideline-bound template, validates safety, and logs the outcome.

1

Classification

Each message is classified by intent (education, symptom, dosing, emergency) and risk level before any response is generated.

2

Template Selection

The right response template is selected from a curated, organization-tunable library — never improvised from scratch.

3

Guideline-Bound Generation

The response is generated within the boundaries of ADA 2025, AHA/ACC 2024, and GINA 2024 clinical guidelines.

4

Safety Validation

Two layers of validators check for dosing language, diagnostic claims, and emergency triggers — blocking unsafe outputs.

5

Audit Logging

Every classification, template, validation result, and final response is logged for organizational review.

Result: 100% governed

No raw LLM output ever reaches a patient. Every response is shaped, checked, and recorded.

Clinical Foundations

Built on the standards your clinicians already follow.

ADA 2025

American Diabetes Association Standards of Care

AHA / ACC 2024

American Heart Association hypertension guidelines

GINA 2024

Global Initiative for Asthma management strategy

Want the full governance walkthrough?

We'll show you live examples of how Dana handles edge cases — dosing requests, emergencies, off-scope questions, and more.

Request a Walkthrough