School-district security sees teachers installing classroom AI agents without review. What procedural control should be embedded?

Embedding an AI change-management procedure that requires assessment and approval before classroom agents are installed is the procedural control school-district security should use when teachers self-install agents. Classroom agents alter data flows, student interaction patterns, and possibly grading or behavioral nudges; change management forces privacy, security, and pedagogical review before installation rather than after a breach or parent complaint. Treating AI agents as changes aligns educational AI with existing ITSM discipline.

A rule that any free browser extension is pre-approved for student use fails because “free” does not mean safe, compliant, or pedagogically sound—and pre-approval without assessment is an anti-control. A ban on documenting classroom AI so auditors cannot find it fails conceptually and ethically: concealment defeats oversight and often violates records or accountability expectations. Delegation of all change approval to students’ personal devices policies only fails because classroom-directed tools are institutional changes, not solely BYOD lifestyle choices.

Exam caveat: change procedures should be lightweight enough for teachers to follow (clear intake, SLAs) or shadow installs will continue; friction without enablement backfires.

Operational check: publish an AI agent change procedure, require assessment/approval before install, remove blanket free-extension pre-approvals, and inventory classrooms for unsanctioned agents.