What is recorded
Analysts rate a verdict with the thumbs on the alert page, and the platform also compares each verdict with the disposition the alert was closed with. Each record holds what the AI said, what the person decided, and on which rule and organization. That record drives:
- the agreement figure for each rule on the Detections page, so you can see where the advice is trustworthy and where it is not;
- guidance drawn from recent decisions, so the advice on a rule improves as your analysts work it;
- the automation gate: when agreement on a rule is low, auto-close and auto-escalate pause for that rule.
Your control
- Feedback stays in your deployment and improves the advice for your organizations only.
- People keep the decision. Feedback sharpens the advice.
Keeping it honest
Feedback only helps when dispositions are honest. Close as benign when it was benign, not when you are tired. The Detections page shows the false-positive rate and the agreement rate side by side, so you can see at a glance which rules need tuning.