Not sure I agree with either/or. In person assessments are still pretty robust. I think an ideal university will teach both with a clear division between them (e.g. whether a particular assessment or module allows AI). What I'm currently struggling with is how to design an assessment in which the student is allowed to use AI - how do I actually assess it? Where should the bar actually be? Can it be relative to peers? Does this reward students willing to pay for more advanced AI?