Based on my experience working on a couple different online homework systems, it's actually a hard problem to determine what kind of wrong answers belong to what kind of conceptual difficulty, even in mathematics.

These systems rely on humans to curate the conceptual difficulties associated with each problem separately, but those lists are never complete and often just confuse the student further.

Some systems don't even try to do what you're suggesting, and just give students the "right" way to solve the problem at hand after a certain number of attempts. These are by far the worst, as it's painfully easy to game the system.

Even in math, this isn't really the case. THere are numerous reasons a student might arrive at an incorrect answer. I've seen systems that provide specific feedback on certain anticipated errors, but tailoring the remediation to individual students is just as hard as teaching the material successfully in the first place. It's often not just a case of "oh, I did this one step wrong", but a deeper issue of misunderstanding the abstraction. We haven't mastered bridging that gap with real-life teachers and tutors, let alone automated software.

It's probably possible. With the leadership team they have, they won't be up to it.