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

# Signal Rules

> Turn customer signals — health drops, negative calls, usage changes, renewals — into AI-drafted recommendations, per segment, with a sensitivity dial and dry-run testing.

Signal rules decide *when* Quivly drafts a [recommendation](/product/actions) for an account. You configure them under **Settings → Signals**.

A rule is like a tripwire per segment: when a chosen signal fires for a customer, the AI drafts next steps and drops them in the Actions inbox.

## Setting up a rule

The editor walks through four steps:

<Steps>
  <Step title="Choose segments">
    Target all customers or a specific segment.
  </Step>

  <Step title="Set up triggers">
    Pick which signals fire the rule: health bucket changes, score drops or rises, negative calls, negative Slack conversations, approaching renewals, usage drops or surges, negative market news, hiring signals, or a semantic "something was said" trigger. Each trigger is an editable plain-English sentence with a Risk / Growth / Neutral polarity.
  </Step>

  <Step title="Configure actions">
    Choose which connected integrations the AI may draft actions through, who suggested actions run as (the CSM owner or the acting user), and optional freeform guidance for the AI.
  </Step>

  <Step title="Configure notifications">
    Optionally notify Slack — DM the account owner or post to a channel, with a fallback channel.
  </Step>
</Steps>

A **sensitivity dial** (Conservative / Balanced / Aggressive / Custom) sets thresholds and cadence in one move, and a quiet-period setting ("stay quiet after handling for N days") stops repeat firing on the same account.

## Testing and publishing

Rules are drafts until published; each publish creates a version. The **Test** tab runs a dry-run preview: how many customers would fire right now, with a sample list and which triggers matched. Semantic triggers aren't evaluated in dry-run.

You can pause action generation, rename, or archive a rule from its settings panel.

## FAQ

<AccordionGroup>
  <Accordion title="How is this different from an agent?">
    Signal rules only produce drafted recommendations for humans to review. [Agents](/product/agents) run multi-step workflows that can act on their own (with optional review gates).
  </Accordion>

  <Accordion title="Will a rule spam me on every score wobble?">
    The sensitivity dial and quiet period control that — after a recommendation is handled, the rule stays quiet for that account for the window you set (1–90 days).
  </Accordion>
</AccordionGroup>