Skip to main content
Use Generate in Email Mode to create a complete ticket-aware draft inside the Zendesk composer. Generate is the main agent-initiated AI Assistant workflow. It can combine the Zendesk conversation with applicable procedures, AI Guidelines, knowledge sources, live SOURCE data, and approved ACTION behavior.

​
Generate a reply

1

Open a ticket

Open the Zendesk ticket you want to answer. Confirm the RAUM mode badge shows Email Mode.
2

Click Generate

RAUM starts processing the current ticket context. Generate works even when the composer is empty.
3

Watch the draft stream

Supported responses stream into the composer while RAUM works.Other conflicting controls are disabled during the active request.
4

Respond to any approval prompt

If a procedure, ACTION tool, or Zendesk ticket change requires agent approval, RAUM pauses and displays the required choice.
5

Review the result

Check the facts, proposed ticket changes, attachments, and wording before sending the Zendesk reply.

​
What Generate can use

Depending on your organization configuration and the ticket, Generate can use:
  • ticket subject and conversation history;
  • requester/channel/Zendesk metadata;
  • KNOWLEDGE and RUNBOOK procedures;
  • Response AI Guidelines;
  • Help Center Articles;
  • Macros and Ticket Field guidance;
  • uploaded documents and websites;
  • active Updates;
  • Gallery assets;
  • SOURCE connections; and
  • ACTION tools when the workflow permits them.
Not every Generate request uses every source. Retrieval and targeting decide which resources apply to the ticket.

​
Cached and pre-cached replies

Administrators can enable Response Caching and Copilot Pre-cache in AI Assistant → Zendesk → Automation → Performance. For an ordinary valid response-cache hit, the app can present a cached-response panel. Choose:
  • Use cached to accept the prepared result; or
  • Regenerate to run a fresh Assistant request.
Copilot pre-cache is different: when the webhook has already prepared a valid pre-cached answer for the ticket, Generate returns that prepared result immediately instead of interrupting the agent with the normal cache-choice prompt. Pending-decision pre-cache entries can also be retained so an agent can resume the decision flow without forcing a full generation again.

​
Stop an active request

A floating Stop button appears while supported generation is in progress. Use it when:
  • you opened Generate on the wrong ticket;
  • the request is no longer needed; or
  • you want to cancel rather than wait for completion.
Cancelling an open approval prompt also ends that pending request without making a decision.

​
Attachments returned by RAUM

A Generate result can reference approved Gallery assets or external attachments selected by the Assistant workflow. RAUM can upload those attachments into Zendesk and associate them with the response when the configured workflow supports it. The organization’s attachment-processing/size settings still apply.

​
Confidence and the sidebar

Generate can return calibrated confidence plus ticket enrichment such as sentiment, category, and summary. After processing, Ticket Insights updates with the available data and can show which procedures/articles/macros were used.

​
Pipeline failures

If the backend marks the AI pipeline as failed, the Zendesk app does not insert the failure text into the composer as if it were a customer response. The ticket can receive the raum_failed tag so it can be routed for manual review.

​
Generate versus Chat Reply

For a realtime Zendesk messaging conversation, use Chat Reply in Chat Mode instead. Chat Reply uses a latency-optimized path and produces a shorter live-conversation response.

Approvals and decisions

Understand why Generate may pause and how to continue.

Ticket Insights

Review confidence, enrichment, feedback, and sources used.