Conference highlights
The three pillars for reliable agentic quality
What QA agent design looks like when patient safety is on the line
Qualifacts builds electronic health records for mental health and addiction treatment, where patient safety is the primary KPI — ahead of shipping software, ahead of revenue.
Adam Satterfield, Vice President of Engineering for Quality and Delivery, explains the three pillars his team applies to every QA and dev agent it builds. A judge agent reviews the first agent’s output, since a QA agent should never be its own final approver. A shared context layer gives QA agents the same evidence the dev team works from. And a human expert retains approval over anything touching patient safety.
“AI could be extremely successful if we just learn to manage risk.”
If you work in a regulated industry and need agents that prioritize traceability, auditability, and safety, this is a good foundation to start from. Watch to learn:
- How to decide which AI outputs still need human review before they ship
- How a separate judge agent lifted one QA agent’s acceptance rate by 50-60%
- What it takes for your AI’s context and output to hold up under audit