I agree that complex questions are the real test for any AI support tool. The ability to understand context, combine information from different sources, and take action makes a much bigger difference than simply answering FAQs. I also think the fallback to human support is just as important as the AI’s ability to solve the issue. Accuracy, good knowledge sources, and clear escalation rules ultimately determine whether the AI actually reduces support workload.