Not sure, those require less talking to machines and more talking to humans. I think jobs talking the most to machines should be easier to automate first in the future, because they obey to logic. LLM doesn’t follow that idea, but that’s just the latest mediatic model, there are many other algorithms better at rational tasks.