AI should improve support and relevance without becoming an opaque decision-maker. Every AI capability in the platform is designed to be explainable, grounded and reviewable by a competent human.
What AI is allowed to do
- Recommend a pathway using approved inputs, explaining its rationale — and always allowing human correction.
- Offer study assistance grounded only in approved course sources, citing the source and stating uncertainty.
- Suggest formative feedback, while high-stakes grading stays governed and reviewable.
- Rank mentor matches on declared criteria, with an authorised person confirming the match.
What AI is not allowed to do
- No fully automated admission, exclusion, certification or disciplinary decisions.
- No training of external models on learner data without an explicit legal basis.
- No sensitive-trait inference or behavioural manipulation.
- No unsupported claims that generated answers are authoritative, unbiased or confidential.
A release checklist, every time
Before any AI capability ships, it must pass a responsible-AI checklist covering purpose, data, accuracy, fairness, transparency and human oversight. Users are told when AI is involved and how to challenge an outcome — and a competent human can always review, correct, stop and escalate.
This article explains our approach and methodology. Impact figures and case studies are published with their evidence and verification level as the pilot produces them — see Impact.