Conversation assistant
The conversation assistant uses the chatt2.me Hub's artificial intelligence to give the agent context during service. It is triggered by the sparkle icon in the composer.
What it delivers
- Conversation summary — what the customer wants, what has already been said, and what remains pending.
- Mood thermometer — a visual indicator of the customer's tone throughout the conversation, shown in the contact panel.
- Reply suggestions — ready-made alternatives for the agent to approve, edit, or discard in the bar above the composer.
The result appears after a few seconds, without taking the agent out of the conversation. Nothing is sent automatically: every suggestion goes through the approval of whoever is handling the service.
How it works inside
- The call goes to the hub's
/assistendpoint, authenticated by the conversation's channel. - The hub uses the session history to generate the summary and suggestions; agents2.me only displays the result.
- The CX Score of a finished conversation also depends on the conversation having had an AI session in the hub — without a session, there is nothing to evaluate.
info
If CX Score is turned off for the company, the evaluation becomes unavailable (cx_disabled) and must be enabled by an administrator in the company settings.
When to use
| Moment | Recommended use |
|---|---|
| Long conversation, before replying | Read the summary instead of rereading everything |
| Angry customer | Look at the thermometer before writing |
| New agent | Use the suggestions as a starting point |
| Before transferring | Generate the summary and paste the note into the transfer |
| End of service | Check whether anything is still pending |
Limits
- The suggestions follow the same permissions as the logged-in user: an agent only generates what they can see.
- Each call consumes resources from the hub's AI provider; usage appears in the hub's consumption reports.
- The assistant does not replace service: it informs, and the agent decides.