Canvas rolled out an AI teaching agent this week designed to automate low-value instructional tasks while stopping short of fully automating grading, and the University of Sydney published a menu-based assessment approach that permits AI use outside supervised classrooms. Canvas framed its agent as a faculty time-saver for administrative tasks; Sydney’s model lets instructors offer assignment options and accept AI assistance for out-of-class work while preserving in-class assessment integrity. Both moves reflect pragmatic campus responses to widespread student AI use: vendors and institutions are piloting tools and assessment designs to preserve learning outcomes while managing workload. Academic leaders, assessment directors and instructional designers should evaluate tool governance, FERPA/privacy safeguards, and rubric redesigns to ensure validity and academic integrity as agentic AI becomes commonplace.