Skip to main content
Open AI Assistant → Zendesk → Automation to control when Assistant runs and how automated work is delivered. Automation contains three areas: AI Mode, Webhooks, and Performance.

​
AI Mode

Choose:
  • Copilot Mode for agent-requested, agent-reviewed responses; or
  • Agent Mode for webhook-driven autonomous processing.
Agent Mode additionally lets you configure:
  • internal-note-only delivery;
  • public reply confidence bands;
  • fallback ticket completion; and
  • required-field completion.
Agent Mode →

​
Webhooks

The webhook settings control which Zendesk events can invoke RAUM’s automated behavior. The onboarding flow normally creates/reconciles the standard webhook and triggers first. The configuration can also produce trigger payloads for enabled Assistant features where manual Zendesk setup is required.

​
AI skip rules

Skip rules prevent selected tickets from entering AI handling. Supported rule types include:
  • phrases;
  • regular expressions;
  • Zendesk channels;
  • tags;
  • Zendesk organizations; and
  • Zendesk groups.
Use skip rules for tickets RAUM should never process, not as a late-stage instruction hidden inside a procedure.

​
Response caching

Response Caching can reuse a valid recent result. Administrators can:
  • enable/disable response caching; and
  • set the cache duration from 1 to 480 minutes.
Clear the organization cache after major knowledge/model/configuration changes when you specifically need fresh results immediately.

​
Copilot pre-cache

Copilot Mode can pre-generate answers when qualifying new-ticket webhook events arrive. Configure:
  • whether pre-cache is enabled;
  • a 1–48 hour pre-cache TTL; and
  • optional Zendesk channel restrictions.
If no channels are selected, pre-cache applies to all supported channels. Pre-cache exists to improve agent latency. It does not mean every later Generate operation must reuse the cached answer regardless of context.

​
Clear all cache

The Cache Management control clears organization cache including pre-cached ticket responses, response cache, and tool-result cache. Use it deliberately because clearing everything trades freshness for increased provider/tool work on subsequent requests.

Copilot Mode

Understand agent-reviewed generation and pre-cache.

Agent Mode

Configure confidence-based autonomous delivery.