Finally circling back around to this.

Feels like I am not just doing my job but also the work the operator of the service or product I am having to use through chat should have paid professionals to do. And I’m not getting paid for it.

Speaking as someone who has worked extensively in IT support, I think that’s the sales pitch for these chatbots. They don’t want to give users tools and knowledge to solve their own problems - or rather they do but the chatbots aren’t part of that. The chatbots are supposed to replace the people who would interact with the relevant systems on your behalf. And honestly, working with a support person is already a deeply unsatisfying interaction in the vast majority of cases. In even the best case scenario it involves acknowledging that some part of your job has exceeded your ability and you need specialized help, and handling that well is a very rare personality trait. But the massive variety of interconnected systems that we rely on are too complex for this to not be a common occurrence. Even if you did radically open everything from internal bug trackers to licensing systems to communications there wouldn’t be enough time in the day for everyone to learn those systems well enough to perfectly self-solve all their problems, and that lack of systems knowledge would be a massive drain on your operations. But trying to fit in an LLM chatbot is the worst of both worlds, in that your users are both locked away from the tools and knowledge that would let them solve their own issues but still need to learn how to wrangle your intermediary system, and that system doesn’t have the human ability to connect and build a working relationship and get through those issues in a positive way.