The thing about AI up to this point is that it has replaced, well, very few jobs, but what jobs it seems to be most capable of replacing are extremely counter-intuitive. Eight years ago, everyone thought it would be: self-driving, machines, hard labor, etc. Turns out that stuff is really hard, and the first industries to fall were actually the more creative ones like writing and art.
If you have an intuition for how the next ten years will look, I'd implore you to be open to the reality that there is no way you could have predicted the world-state of 2024 from the perspective of 2019; and that's only 5 years.
Here's the counter-intuitive take that I believe to be true: Specialized coding might fall faster than generalist CRUD apps, if either falls at all. The value in specialized engineering is biased a lot heavier toward knowledge of the specialized thing you know about. Versus, the problems generalist CRUD coders deal with every day aren't actually technical or coding problems; they're business problems, coordination, resource allocation, and politics. AI has demonstrated itself as being pretty good at knowing things, even highly specialized things; it has not demonstrated itself as being very good at taking responsibility for its actions.