M8-B — Internal Knowledge Systems

🏢 What is an internal knowledge system?

It is the company's shared brain on disk:

SOPs + policies + training docs + past project notes
                    ↓
            Searchable with AI
                    ↓
    "How do we handle refund requests?"

Brochure wording: "Internal knowledge systems" and "Building document-based AI assistants" — same idea, bigger audience.

😫 The problem it solves

This is the hidden tax from M6-A — AI for Decision-Making and Business Productivity — now you fix it with RAG.

📦 What goes inside the knowledge base

🔄 How teams use it daily

Morning: Staff member unsure of policy
    ↓
Opens internal chat (Notion AI / custom / NotebookLM shared)
    ↓
"As per our refund policy, can we refund after 7 days?"
    ↓
AI pulls refund SOP chunk → answer with citation
    ↓
Staff confirms with manager if edge case

Human still owns final decision on exceptions.

🛠️ Tool options (2026 landscape — samples change)

Tools update fast — the pattern (upload → index → chat → cite) stays the same.

⚙️ Architecture (team version)

        ADMIN uploads/updates docs
                    ↓
            ┌───────────────┐
            │ Knowledge base │  ← vector index refreshes
            └───────┬───────┘
                    ↓
    ┌───────────────┼───────────────┐
    ↓               ↓               ↓
 Staff chat    Student portal    WhatsApp bot (later)

Backend jobs you don't see:

• Re-index when PDF updated
• Log questions people ask (find gaps in docs)
• Block answers outside allowed folders

🎯 Brochure capstone tracks

Same RAG engine. Different documents. Different personality prompt.

✅ Build order (recommended)

1. Pick one audience (students OR staff)
2. Collect 5 core PDFs
3. Test in NotebookLM for 1 week
4. Write personality + refusal rules
5. Only then think about custom website or WhatsApp integration

Previous: M8-A — Document-Based AI Assistants
Next: M9-A — What Are AI Agents — when the system starts doing tasks, not just answering