Rovo AgentResponse Confidence Score

[Ramesh Muthusamy]

Hi Rovo community — question on using confidence signals for customer-visible comments.

We have a Rovo agent that triggers when a new case is created. It searches for similar previously resolved cases in the JIRA project and/or relevant Confluence articles, then posts the suggested solution as an internal comment.

We’d like to extend this so the agent can post the response as a requestor-visible comment only when the answer is highly likely to be correct. Otherwise, it should remain an internal comment.

Is there a way to access or determine a confidence score for the AgentResponse (or any reliability signal/metadata)?
Example logic:

• If confidence > 90% → post as requestor-visible comment

• Else → post as internal comment

Any guidance on whether this is supported (and how to implement it) would be appreciated.

[Ryan Boyd · Accepted solution]

I'd have the agent generate the confidence score, write that score to the ticket, and then use automation to fork the response for >90% to the user and <90% internal. That *should* work in my experience. 

[Dr Valeri Colon _Connect Centric_]

Welcome to the community @Ramesh Muthusamy I'll add to the suggestion by saying: 

Rovo doesn’t expose a confidence score or reliability metadata for agent responses, so you can’t automate “only post externally if confidence > X.” The agent output is text-only, and Automation receives no confidence field to evaluate. Current patterns are: post internally first, have the agent label its own certainty in the text (not guaranteed), or route through a human-review step. True confidence scoring isn’t supported yet.