The ticket copilot (Assist)

The copilot sits beside every ticket in the workspace. An agent hands it a task in plain language — “draft a reply asking for the exact error message”, “summarize this thread for the escalation” — and it does the work inside that ticket’s real context.

What it can do

Depending on the actions your organization allows:

• Reply to the requester — drafts (or sends, if allowed) a public reply.
• Add internal notes — analysis, summaries, next steps, visible to your team only.
• Update the ticket — status and fields, when allowed.
• Answer questions about the ticket’s history and related knowledge.

Every action is recorded in the ticket’s history under the AI’s name — nothing it does is anonymous.

Allowed actions are your call

1. Open Organization → Settings → AI → Assist.
2. Enable or disable each action the copilot may take, and whether replies require approval.
3. Save.

The policy is enforced server-side: an action the policy forbids is refused even if requested, and the agent sees why.

Recent actions

The copilot panel shows its recent actions on the ticket — what was asked, what it did, when — so a colleague picking up the ticket sees at a glance what the AI already handled.

Voice input

The task box accepts dictation in supported browsers — speak the instruction instead of typing it. Interim text appears as you speak; review before sending.

The same grounding rules

The copilot uses the same retrieval as the email agent: your knowledge base and data sources, with visibility respected. What it cannot ground it flags, rather than inventing.