Skip to content
SmartChatSmarChat Docs
How it works2 min read

Confidence Level: automatic responses and suggestions

How the Confidence Level decides whether the AI sends a response on its own or proposes it as a suggestion, how to recognize the two cases in the chat, and how to use, modify, or analyze a suggestion.

This page is not translated yet: you are seeing the Italian version.

You are in the right place if…

  • “The AI does not respond on its own, it only offers me a suggestion”
  • “What does the green box in the chat mean?”
  • “The AI sends too many responses without asking me anything”
  • “Why did the AI suggest this response?”

This guide has not yet been verified against the latest software version.

Every response prepared by the AI receives a reliability score from 0 to 5. The Confidence Level set in the AI Profile is the threshold: if the score is equal to or higher, the response is sent automatically; otherwise, it remains a suggestion to be approved.

  • Low threshold: the AI sends messages on its own more often.
  • High threshold: the AI is more cautious and proposes more suggestions.
  • Never send: the AI does not send anything on its own and only proposes suggestions.

Example: with a Confidence Level of 3, responses with a score of 3, 4, or 5 are sent automatically; the others remain as suggestions.

How to recognize them in the chat

  • Light blue message with the AI symbol: response sent automatically.
  • Green box: suggestion not sent. You can click Use to send it as is, Edit to bring it into the writing area and correct it, or info to see why the AI proposed it and from which document or rule it took the information.

If the suggestion does not convince you, write your own response in the box below.

Which level to choose

At the beginning, it is advisable to use a cautious threshold, or "Never send", for a few days: you check the suggestions, complete the Knowledge Base where something is missing, and then lower the threshold once the responses are reliable.

Updated on October 6, 2026