Along the way, I've done what I think a lot of AI power users eventually wind up doing: I've gone into the personalization and settings and told the chatbot to be neutral, direct, and just-the-facts. I don't want a chatbot that tells me "That is a brilliant idea!" every time I explore a tweak to my business strategy. They're not all brilliant, I assure you. And I don't want a lecture about how if I truly have shoulder issues I should see a "real" physical therapist. I'm an adult. I'm not outsourcing my judgment to a robot.
There were specialists monitoring dashboards, tuning AI behavior, debugging API failures, and iterating on knowledge workflows. One team member who had started their career handling customer questions over chat and email (resetting passwords, explaining features, troubleshooting one-off issues, and escalating bugs) was now writing Python scripts to automate routing. Another was building quality-scoring models for the company's AI agent. This seemed markedly different from the hyperbole I'd been hearing about customer support roles going away in large part due to AI.