Feedback on AI chat quality has nowhere to go. Admins can test bot responses in the chat testing environment and can review real conversations after the fact, but in neither case does "this answer was wrong" change anything. The only lever is manually rewriting guidance and hoping it covers the case, with no record of what went wrong or why. Requested behavior: let admins label a response or a whole conversation as a correct or incorrect outcome and attach a rationale explaining what the AI should have done instead. Labels should be available both in the testing environment and on live production conversations, and the labeled examples should feed into how the AI answers and routes going forward, so the same mistake stops repeating. Why it matters: teams are auditing conversations manually and translating what they find into guidance edits by hand, which is slow and loses the reasoning behind every correction. A conversation-level labeling loop turns real chats into the training signal, shortens the time between spotting a bad outcome and fixing it, and gives teams a defensible record of how AI behavior was corrected over time.
