We’d love to see QA summaries surfaced directly in Assembled, ideally on the AI Performance page. This would help us quickly determine whether a poor interaction was caused by a rogue workflow, a knowledge base gap, or an agent issue, without having to leave Assembled.

Today, our QA team manually reviews approximately 1,000 AI chat and email interactions and then exports QA scores from the Quality Review page, runs them through an LLM to generate a summary, and then manually analyzes the comments to identify recurring issues.