LLMs aren't perfect, but I would vastly prefer to be assisted by an LLM over the braindead customer service chatbots we had before. The solution isn't "don't use LLMs for this," but instead "take what the LLMs say with a grain of salt."
LLM’s are still in their infancy and easily mislead with the right prompting, and are still far too prone to hallucination to have applicability in the way some people are trying to implement them.