Intelligence is the sixth pillar of the Nomotic AMP. It learns from your agents’ behavioral history and surfaces insights that help you improve governance over time.
Recommendations
The Recommendations tab shows findings generated by Nomotic’s analysis engine. Each finding identifies a pattern in agent behavior that warrants attention, with a suggested action.
Finding types include:
Verdict drift — your agent’s allow rate changed more than 20% week over week.
UCS decline — average confidence scores are trending down over the past week.
Idle agent — an active agent has not been evaluated in 14 or more days.
Cost outlier — this agent’s cost per run is significantly above your fleet average.
Missing context — this agent has evaluations but no context profiles assigned.
Escalation pattern — the same action type is escalating repeatedly.
Each finding shows a severity level (low, medium, or high) and a suggested action. You can approve the suggestion (Nomotic applies the recommended change) or dismiss it. All findings expire after 30 days if not acted on.
Analytics
The Analytics tab shows fleet-wide performance over time:
Verdict distribution — how ALLOW, ESCALATE, and DENY verdicts are trending across your fleet.
UCS trend — average confidence scores per day, per agent, and fleet-wide.
Escalation ROI — of your escalations, how many resulted in human approvals versus denials. This helps you tune escalation thresholds.
Governance coverage — what percentage of your active agents received at least one evaluation this week.
Cost tracking — evaluation cost per agent and fleet total over time.
Fleet Context
The Fleet Context tab shows the Behavioral Context Window for every agent in your organization — their baselines, trends, anomaly scores, and dominant action patterns — in one view. This is where you spot behavioral changes across your fleet before they become problems.