AI won't make the analyst skillset obsolete. You need to be a player before being a coach

Obviously difficult to know how the future plays out, but wanted to hear people's thoughts on this argument: I do not think a day will come where an analyst doesn't need to learn modeling, slide constructing, or even basic admin tasks from scratch. 

Here's why: When you actually do a task yourself, you learn at a completely different depth than you do when you watch someone else do it then go back and check their work. When you do something yourself, you understand the thought process step by step and where errors are more likely to occur - why they are so, most importantly. How often do you watch a lecture and everything makes sense, then when confronted with a blank sheet and the same problem, you suddenly aren't so sure exactly what to do? Or when watching a sports game, how much easier is it to criticize a professional athlete (e.g. they should have passed to the guy on their left or taken a shot instead of dribbling), as opposed to being out there on the field yourself? Only when you're "in the field" and seeing everything from their POV, can you then see why they did what they did at the time.

Is it fair to say this analogy might apply to AI? I understand that AIs do not process information the same as we do, but I would imagine that they may be susceptible to similar errors. How can you expect someone to catch these if they have not been "in the field" themselves at some point? Because of this, I don't really see raw financial modeling skills and analyst tasks ever becoming obsolete. However I can see a world where you are asked to reach a level of proficiency, then check AI's work on a day to day