It depends on the objective.

Make a graph with two axes.

Vertical is longevity. Throw-away apps at the bottom, software meant to last decades at the top.

Horizontal is how critical the software is. Ad-hoc or experimental or trivial stuff near the origin, financial, medical, or other important software that must never make a mistake to the far right.

AI can sit in the lower left. It is meant to do stuff in the lower left. Let it go nuts there, and have fun with it.

But AI cannot be allowed to touch the top right corner. Up there is the region of destroyed lives if something goes wrong. Sure, rubber-ducky things with AI, let it read the code in that corner, let it suggest optimal structures for you to build. But never let it touch code that’s sitting there.

As for the other two quadrants… “it depends” on your appetite for risk, of AI going so badly off-script that the company is destroyed trying to untangle that hairball. YMMV.