The Only Medical AI with a Supervisory Agent
POPPER ensures clinical safety at every step. AI proposes. Clinicians decide. Every recommendation passes through safety evaluation before reaching patients or providers.
A dedicated supervisory agent that watches every AI recommendation in real-time.
Audit trail for every AI decision
Autonomous AI actions without review
POPPER safety monitoring
Why POPPER Matters for Your Organization
Named after philosopher Karl Popper, our supervisory agent applies the same rigorous falsification principles to every AI recommendation - ensuring nothing reaches patients without passing safety evaluation.
Risk Mitigation
AI never acts autonomously. Every recommendation is a proposal that passes through POPPER's safety evaluation before reaching clinicians or patients.
Liability Protection
Complete audit trail for every AI decision. Know exactly what was proposed, what was approved, and who made the final call.
Clinician-in-the-Loop
AI proposes, doctors decide. Your clinical staff maintains full authority over all patient care decisions. AI is a tool, not a replacement.
Automatic Escalation
When AI is uncertain, it says "I don't know" and routes the case to appropriate specialists. No guessing, no false confidence.
No Other Medical AI Has This
While other AI systems operate as black boxes, Regain is the only company with POPPER - a dedicated supervisory agent that watches every recommendation in real-time. This is your competitive advantage in clinical safety.
Three Agents Working Together
Our agentic AI system is built on three specialized agents, each with a distinct role in ensuring safe, accurate clinical recommendations.
DEUTSCH
Knowledge Engine
- Generates multiple hypotheses
- Verifies against evidence
- Reasons through findings
- Scores reliability (HTV)
HERMES
Message Contract
- Structures all data
- Ensures consistency
- Validates format
- Enables audit trail
POPPER
Safety Referee
- APPROVE safe actions
- ROUTE uncertain cases
- HARD STOP risky actions
- 24/7 monitoring
How the Agents Work Together
DEUTSCH
Creates knowledge
HERMES
Structures data
POPPER
Evaluates safety
Clinician
Makes decision
Built-In Safety Guarantees
Every recommendation passes through multiple safety layers. Here's how we ensure clinical safety at every step.
HTV Scoring
Hard-to-Vary Threshold
Every recommendation is scored based on how "hard to vary" it is - meaning how well it's supported by evidence and how difficult it would be to construct an alternative explanation.
Only high HTV scores pass safety review
IDK Protocol
7 Triggers for Uncertainty
When AI detects any of these 7 conditions, it immediately says "I don't know" and routes to a human clinician:
- Conflicting evidence
- Insufficient data
- Novel presentation
- High-stakes decision
- Edge case detected
- Model uncertainty high
- Outside training scope
Feedback Loop
Confidence Decay System
When clinicians override AI recommendations, the system learns. Confidence scores automatically decay for similar future cases until validated.
Clinician Override
→ Triggers confidence decay
POPPER Decision Matrix
| Decision | Condition | Action |
|---|---|---|
| APPROVE | High HTV score, no IDK triggers, within scope | Proceed to clinician review |
| ROUTE | Medium confidence, some uncertainty flags | Route to specialist for additional review |
| HARD STOP | Low confidence, multiple IDK triggers, safety concern | Block action, require manual clinician intervention |
Ready for Safer Medical AI?
See how Regain's POPPER supervisory agent can reduce liability, increase trust, and bring your organization the safest medical AI on the market.