September begins.
And with it the idea that the academic year is a starting point. It is not. September is the month when decisions get frozen that can no longer be touched until next year. This year there are three on the table, and only one of them is being discussed.
1. AI is not an honesty problem. The test has expired.
Everyone has quoted the same finding this summer: students who study with AI do better in practice and worse in the exam without a computer. The conclusion almost always drawn is that AI is taking learning away from them.
I think it measures something else. Ban an engineer from simulation software and make them calculate by hand, and they will score worse too, yet nobody would say they know less engineering. You are measuring the restriction, not the competence.
So September's challenge is not deciding what you allow. It is accepting that if your assessment breaks the moment a model shows up, the problem is the assessment. And then the question becomes uncomfortable: which part of this course is still hard with an AI in the room? Whatever survives that question is the syllabus. The rest is decoration, and this year it will show.
2. The evidence you will be asked for in three years is decided now.
AACSB and EQUIS do not ask for activity; they ask for evidence of learning. What most institutions have on file is activity: attendance, submissions, satisfaction scores and a rubric filled in by a single person in the last week.
The difference between the two does not surface until the visit, and by then that cohort has already graduated. A course that has already run cannot be instrumented after the fact. September is the only moment in the year when instrumenting it costs almost nothing, because the design is still open.
3. The bottleneck is not technology. It is faculty hours.
Almost every institution we work with has more tools under contract than it uses. The scarce resource was never the licence: it is the hours of a professor with judgement who has to redesign a session and then defend it in front of colleagues.
Those hours appear in no budget. The purchase is budgeted and the redesign is taken for granted. That is where adoption dies, not at the buying decision.
All three are the same problem in three different places: for years we have been measuring what is easy to collect instead of what we meant to teach. AI did not create that problem. It has only made the bill arrive sooner.
I have spent ten years working on how people are assessed, and I still have not solved the third one. It is the least discussed, and the one that has most often sunk a project that looked fine on paper.
Of the three, which one was left out of your course plan — and who decided to leave it out?
