That's the previous status quo; you'd also find this in call centres where customer support had to follow scripts, essentially as if they were computers themselves.

Even quite a lot of new chatbots are still in that paradigm, and… well, given the recent news about chatbot output being legally binding, it's precisely the extra agency of LLMs over both normal bots and humans following scripts that makes them both interestingly useful and potentially dangerous: https://www.bbc.com/travel/article/20240222-air-canada-chatb...