Human takeover
When a workspace has the AI assistant turned on, the bot answers customers automatically. Human takeover is how a real agent steps in and how the bot steps back out — so a person and the AI never talk over each other in the same conversation.
Each conversation has its own AI switch. While it is on, the bot may reply. Once a conversation is handed to a human, the switch turns off and the bot goes quiet for that conversation only — every other conversation keeps running on AI as before.
When the bot hands a conversation to a human
A conversation is handed over in three situations:
- The customer asks for a person. In the chat widget the visitor taps Talk to an agent, or simply writes something like "I want to speak to a human" / "let me talk to an agent". The bot recognises the request (in English and Turkish), stops replying, and flags the conversation for an agent.
- The bot can't help. When the assistant has no relevant answer for a workspace that has a knowledge base, it can escalate to a human instead of guessing. (This automatic escalation is conservative and off by default — your admin can enable it.)
- An agent takes over manually. From the inbox an agent can grab any conversation at any time, even one the bot is actively handling.
In every case the result is the same: the conversation's AI switch turns off, and the conversation is marked waiting for an agent until someone picks it up.
What the customer sees
While the AI is answering a widget chat, the visitor has a Talk to an agent action above the message box (when the handover mode is button). Without AI there is no such button — the visitor is already talking to your team. After they use it (or after the bot escalates) the widget shows:
| State | What it means |
|---|---|
| Connecting you to an agent… | The handover was requested; no agent has joined yet. |
| An agent has joined the chat. | A human is now handling the conversation. |
The customer can keep typing the whole time — nothing is blocked. If they are offline when an agent replies, they'll see the reply the next time they open the widget.
What the agent does
Handover lives in the inbox. Conversations that need attention are easy to spot:
- A conversation waiting for an agent shows a Waiting for agent badge in the list, and a banner across the top of the thread.
- A conversation a human has taken over shows an Agent badge instead of the usual AI indicator.
Open the thread and use the control in its header:
- Take over — assign the conversation to yourself and silence the bot. Use this whenever you want to handle a customer personally. The badge flips to Agent and the bot will not reply again until you hand it back.
- Give back to AI — re-enable the assistant for this conversation. The bot resumes answering the customer's next message. You can do this for any conversation you handle; supervisors can do it for any conversation.
Taking over and giving back are instant for everyone watching — the customer's widget and your teammates' inboxes update live.
Taking over assigns the conversation to you. To move it to a different agent, use the normal assignment controls; handover only governs whether the AI is allowed to reply.
Privacy & isolation
- A visitor can only ever hand over their own conversation — the widget acts on the conversation tied to its secure session, never one chosen by the browser.
- Handover is scoped to a single workspace. One workspace can never see or change another workspace's conversations or AI state.
- Turning AI off for one conversation never affects any other conversation, and never changes your workspace-wide AI settings.