I conduct research in a very specific field. AI can help us along but it is lightyears away (not actually) from replacing what I do. Finding holes in the theory, developing a model, designing a study, analyzing the data (AI can do that to a degree but understanding what analysis to use specifically is difficult for AI) but then interpreting the data including the judgement calls that have to be made via expert opinion, then writing it all up. AI is just too fair off.

For instance, the other day one of my results produced a value of .67 and for this particular metric the cutoff is .70. However, using my expert knowledge and theory I can easily rationalize the .67 as actually being a "success" do to a number of factors. The abstract reasoning required for that assessment is far beyond AI at the present time and instead AI would hold to the .70 and would actual falsely reject the parameter.