AI should be visible, constrained and reviewable.
RCI V13 separates learner-facing guidance, educator assistance, operational automation and analytics so each can have different permissions, safeguards and human oversight.
Learner Guidance
Ask guiding questions, explain concepts and encourage evidence. Do not conceal AI involvement or make clinical/safeguarding decisions.
Educator Assistance
Suggest lesson preparation, resources and draft feedback while keeping educators responsible for final decisions.
Operational Automation
Low-risk administrative triggers can be queued automatically; sensitive or irreversible actions should require approval.
Content Governance
Generated educational content should have provenance, review status and version history before publication.
Data Minimisation
Only necessary learner context should be supplied to an authorised AI service when one is connected.
Human Escalation
Safeguarding, distress, legal, payment disputes and high-stakes educational decisions leave the automated flow.