Escalations are how Nomotic involves humans in governance decisions that require judgment. This article explains when escalations fire, how to manage your approval queue, and how escalation decisions are recorded.
When escalations fire
An escalation fires when an agent action scores in the ambiguous range — high enough that an automatic denial seems too harsh, but not high enough for Nomotic to approve it confidently. The exact threshold is configurable in Governance → Policies.
Common triggers include actions outside the agent’s typical behavioral pattern, actions that score low on the authority verification dimension, actions in a new context the agent hasn’t been seen in before, and actions with potential cascading impact that warrant a human second opinion.
The approval queue
All pending escalations appear in Governance → Approval Queue. Each item shows the agent name, the action it was trying to take, the UCS score, and which dimensions triggered the escalation.
To review an escalation, click it to open the full evaluation detail. You’ll see the complete dimension breakdown, the reasoning artifact, and the agent’s recent behavioral history. From there you can approve the action (it proceeds immediately) or deny it (it is blocked and the agent is notified).
Who can review escalations
Team members with the Admin or Operator role can review and act on escalations. Viewers can see the queue but cannot approve or deny. Escalation routing — which team members are notified — is configurable in Governance → Policies.
Escalation expiry
Escalations expire after the time window configured in your policy (default 24 hours). An expired escalation is treated as a denial — the action is blocked and the expiry is logged. For time-sensitive agents, configure shorter expiry windows and ensure your notification routing reaches someone who can respond quickly.
How escalation decisions feed back into governance
Every approval and denial is recorded alongside the original evaluation. Over time this creates a pattern of human decisions that Nomotic’s Intelligence layer analyzes. If your team is approving a high proportion of escalations for a specific action type, that is a signal that your escalation threshold for that action type is too aggressive. Intelligence will surface this as a recommendation.