Where do we draw the line between AI "merely" polishing the grammar of a text–and it actually evaluating the professional content? And ultimately: is it acceptable for an algorithm to decide whether the work of a colleague or a student is good enough?