AI Agent answer traceability on live tickets

Business problem
My Customer Support team or Sales team cannot trace back where the AI Agent's answer to a customer comes from when handling a live ticket. When the AI Agent gives an incorrect or unexpected answer in production, we have no way to see which scenario, knowledge source, or logic it used to generate that response. This makes it very hard to diagnose why the AI Agent sometimes answers correctly in the test environment ("Try it out") but incorrectly in a real, live customer conversation — we can only guess which scenario was triggered instead of confirming it.

Wished result
I would like to have a debug/trace feature for the AI Agent, similar to the trace functionality that already exists for Journeys, that shows exactly which scenario, knowledge base article, or source was used to generate a specific answer on a live ticket. This would let CSMs and support teams quickly identify the root cause of incorrect AI Agent answers and compare live behavior against the test environment, instead of having to reconstruct it manually.