You claon accuracy. LLMs results typically have dire repeatability i.e. same input gives different output. Hence anyone relying on tests for accuracy is kidding themselves.
You might not know how to do this, which is fine, but it’s not a new problem.
Non-deterministic is one thing, unreliable and inaccurate is quite another. E.g. see monte carlo simulation
Regradless, I didn't suggest chatbot-generated bullsh*t wasn't valuable to businesses. On the contrary. The amazing productivity of these chatbots is every day making it more valuable to all types of bullsh*t-based businesses worldwide.