> ## Documentation Index
> Fetch the complete documentation index at: https://help.craftcx.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Support Observability

> Turn your full stream of AI support activity into a live operational view—spot patterns, detect policy violations, and route findings to the right people before they become bigger problems.

Support Observability is CraftCX's operational layer for AI-powered customer support. Instead of asking your team to hunt through individual conversation scores or dashboards, it continuously extracts signals from incoming support activity, groups related signals into persistent issues, and surfaces a prioritized view of what needs attention.

<Info>
  Support Observability is currently in active development. Features are being rolled out progressively—reach out to [support@craftcx.com](mailto:support@craftcx.com) for early access.
</Info>

## What's included

<CardGroup cols={2}>
  <Card title="Policy Monitoring" icon="shield-check" href="/support-observability/policy-monitoring">
    Continuously verify that your AI support behavior follows your organization's policies, SOPs, and customer commitments. Flag violations with evidence and track them to resolution.
  </Card>

  <Card title="Incidents" icon="triangle-exclamation" href="/support-observability/incidents">
    Repeated or high-severity policy violations are promoted into incidents. Assign ownership, set due dates, and track status through to resolution.
  </Card>

  <Card title="Notifications & Alerts" icon="bell" href="/support-observability/notifications">
    Route findings to the people and systems that need them—via Slack, email, or webhook—as they happen or as a scheduled digest.
  </Card>
</CardGroup>

## How it fits together

CraftCX evaluates every AI-handled conversation using [AXIS Scoring](/features/axis-score) and [Audit](/features/audit). Support Observability builds on top of those signals:

1. **Policy Monitoring** evaluates each conversation against your uploaded policies and SOPs, flagging violations with severity and evidence.
2. Repeated or severe violations are promoted into **Incidents**—persistent objects your team can triage, assign, and resolve.
3. **Notifications** ensure the right people are alerted when a new incident is detected or when AXIS scores drop below a threshold.

This means patterns that affect customers—a product regression, a billing edge case, an AI responding incorrectly to a refund policy—surface quickly, with evidence, rather than going unnoticed in individual tickets.