The email agent
The email agent works your support mailbox. Every email a connected channel receives passes through it; whether it answers is entirely up to your configuration.
Turn it on
Section titled “Turn it on”- Open Organization → Settings → AI.
- Enable the agent and give it a persona: name, display name, tone, language, and an optional signature and custom instructions.
- Choose the response mode and sender policy below.
- Save.
Response modes
Section titled “Response modes”| Mode | Behavior |
|---|---|
| Autonomous | The agent sends its reply itself when confident enough |
| Draft for review | The agent prepares the reply; a human approves and sends |
Start in draft mode, read a week of drafts, then decide. The confidence threshold applies in both: below it, the agent never answers — it leaves the ticket for your team with its analysis as an internal note.
Who the agent answers
Section titled “Who the agent answers”The sender policy decides which senders may receive an AI reply at all. Combine:
- Internal users — people with accounts in your organization.
- Known clients — contacts of your registered customer companies.
- Your domains — anyone writing from your organization’s email domains.
- Everyone — any sender (use deliberately).
- Allow / block lists — specific addresses and domains that override the above.
A sender the policy excludes is never answered by the AI — their email still becomes a ticket and reaches your team normally.
Safety caps
Section titled “Safety caps”- Max replies per conversation — after it, the agent hands off instead of looping.
- Max replies per sender per hour — an auto-responder on the other side cannot create an email storm.
- Unknown senders can be excluded in one switch.
Ticket-aware, not just email-aware
Section titled “Ticket-aware, not just email-aware”The agent reads the ticket’s situation before acting: a follow-up on an already-escalated ticket does not restart the conversation or re-escalate it; a reply to a resolved ticket is recognized as such. Status changes the agent makes are visible in the ticket history under its name.
What the customer sees
Section titled “What the customer sees”A normal email: your branding, the agent’s persona as sender name, the answer, sources from your public knowledge base, the agent’s signature — and, if enabled, a short disclosure that an AI assistant wrote it.
When it does not answer
Section titled “When it does not answer”The ticket stays with your team, carrying the agent’s internal analysis. Nothing is dropped: “the AI stayed silent” always means “a human has it”.