AI will, like previous technologies, enable some of us to become more productive. In fact, it raises the bar on productivity, since an experienced programmer can now create much more code. (An inexperienced one can create much more mess, so you might not see it in aggregate statistics).
When it comes to the classroom, we should do the same. We raise the bar so that in fact, you cannot do anything without using AI. Much as you would run out of time if you didn't have a spreadsheet in a stats course 20 years ago, or pandas 10 years ago. The new tech enables more work to get done in the form of learning more high level things, while relegating lower level things to just building blocks that can be understood in the same way we understand reference texts, ie "I've seen the principal once, and I can find it again if I get to that level of abstraction".
Teaching needs to change. Perhaps the thing to do is have an Oxford tutorial rather than traditional class. For those who didn't attend, a tutorial is basically two students and a professor in a room, talking. You can't hide. You can prepare however you like, and you should spend quite a lot of hours if you're sparring with a politics or math professor. But once you're in the room, it becomes painfully obvious if you are unprepared. This is a way to get accountability.
At the moment, we have this high school system testing that is a factory. Every test is done as a thing that is easily marked. Multiple choice, or short answer, or short essay. It encourages superficial learning when you know you can dance around the important topics and just pick up the easy points, as well as simply avoiding silly errors. You can also win by simply learning the likely questions, and aping the answers.
Have a weekly small-group session with an expert, and they can find your limits. Yes, it will cost money.