The School Knowledge Base (RAG): Letting AI Cite Your Own Materials and Policies

Short answer: a school knowledge base turns AI from "knows about the world" into "knows about your school" — upload curriculum, materials, policies and notice templates, and the AI retrieves, answers and cites. The documents belong to the school, export in full at any time, and are never used to train models.

The blind spot of generic AI

Ask a generic AI "which P4 Chinese unit are we on this term?", "what is the school's homework policy?" or "write a notice in our school format" and it can only guess. It has never read your documents. RAG solves exactly that.

How RAG works (teacher edition)

1. Upload: documents (PDF, Word, slides, web pages) go into the knowledge base; the system splits them into passages and indexes them.
2. Retrieve: when a teacher or student asks, the system finds the most relevant passages.
3. Generate: the AI answers from those passages and appends "Source: Homework Policy 2026, section 3".
4. Trace: users click through to verify the source.

Key point: documents are never used in model training; they are only consulted at answer time.

What belongs in the knowledge base?

Six use cases

1. Planning: "Generate a lesson plan for our P5 Maths unit 6 objectives" — the AI cites the school pacing chart; no re-typing.
2. Tutor: a student asks "what are the key points of this lesson?" — the tutor cites teacher-opened school notes, not random web content.
3. Marking: the AI grades against panel rubrics and cites the criteria in feedback.
4. eAdmin: notices automatically follow school templates and salutations.
5. Staff Q&A: "How do we apply for exam accommodations for SEN students?" — cites the policy with section numbers.
6. New-teacher onboarding: one chat interface answers questions about rules, systems and procedures, with verifiable sources.

Permission design

• School: all staff can retrieve (policies, admin templates).
• Panel: only that subject's teachers (pacing, materials, rubrics).
• Class: materials a teacher opens to a specific class.
• Personal: a teacher's own notes.

Student mode retrieves only teacher-opened class-level documents, behind the safety guard.

Keeping quality high

• Tag every document with subject, grade, school year and owner;
• versioning: when a policy is updated, the old version is kept but no longer retrieved;
• panel heads review outdated documents once a term;
• retrieval logs show which documents are cited most and which questions found nothing (= documents to add).

The data belongs to the school

The knowledge base is the school's accumulated knowledge asset. A qualified platform must let the school:

• export everything at any time (original files + index metadata), free and unlimited;
• delete any document and have the AI stop citing it immediately;
• receive migration help and a deletion certificate at contract end.

A knowledge base grows more valuable with use — which is exactly why it must belong entirely to the school.

Rollout suggestions

1. Month one: only three document types — pacing charts, homework policy, notice templates.
2. Let eAdmin and planning tools use it first so teachers feel the "sounds like our school" difference.
3. Month two: open the panel tier, with panel heads uploading materials and rubrics.

Takeaway

RAG turns AI from a clever outsider into a colleague who knows the school. Documents stay in the school's hands, retrievable at any time, never trained on — the most fundamental difference between a school-based AI platform and a generic tool.