AI Autopilot
Let the agent answer on its own, and take back a single conversation without stopping it everywhere else.
AI Autopilot is the switch that decides whether your agent replies by itself in a given conversation. It is the difference between an agent that assists you and an agent that works for you.
Two switches, two scopes
Autopilot is easy to confuse with the workspace-wide Active switch. They do different jobs:
| Switch | Where it lives | What it controls |
|---|---|---|
| Active | Top of the AI Center | Global. Turning it off stops the agent across the whole workspace. |
| AI Autopilot | In each conversation's bar | Local. The agent answers on its own in this conversation only. |
Take over one conversation
- Open the conversation in the inbox.
- Turn AI Autopilot off in the conversation bar.
- Reply yourself. The writing assistant can still draft for you.
The agent keeps answering every other conversation. Turn Autopilot back on when you are done and it resumes on this one.
When the agent stops on its own
You are not the only one who can pause the agent. It steps back by itself when a conversation is escalated — when a customer is clearly unhappy or explicitly asks for a human.
FAQ and troubleshooting
The agent is not replying even though Autopilot is on
Check the global Active switch in the AI Center first — if that is off, no conversation gets an automatic reply. Then check whether the conversation carries an escalation note, which also stops the agent.
Does turning Autopilot off delete anything?
No. It only stops future automatic replies in that conversation. Nothing already sent changes, and the customer sees nothing.
Should I run with Autopilot on from day one?
Test first. Leave the global Active switch off and use the AI Center's Live Test until you are happy with the answers — see Test and activate your agent.
What's next?
Learn what happens when the agent decides a human is needed.