Evaluating an intent classifier: what I check beyond accuracy

I built the intent router that sits at the front of our support agent. It reads what a user typed and decides which pipeline handles it. I tested it on a nice balanced set, got 92 percent accuracy, an