The Big Picture
You need the Korbay trade licence. Where is it?
You've checked three places and wasted 10 minutes. Sound familiar?
This system ends that. Every file has one permanent address. Every AI worker knows exactly what facts to use. The Memory Card maps all of it.
The 3 Pillars — click each card to flip
The Operating Principle
Every task in this system follows three steps in order. No exceptions.
Key rule: A Gem without a Truth Container is just a chatbot. It will make things up. Always attach the cabinet first.
The Address System
Why Numbers, Not Folders
Drive → Company → Korbay → Finance → 2026 → Q1 → Final → actualfinalv3.xlsx
7 clicks. 30 seconds. Still not sure if it's the right file.
12.06 [A] - Finance - Korbay Master Sheets
16.03 [A] - Operations - Pleck Autos MOT Records
1 scan. 3 seconds. You know exactly what it is.
The Seven-Second Rule: Any person in the organisation should be able to identify any file within 7 seconds from its name alone. If the name fails this test, rename it.
Zero-Folderization: The JD number is the folder. You don't need nested directories anymore.
The JD Block Map
| Block | Domain | Categories |
|---|---|---|
| 00–09 | System Administration | 00 System Index & AI Infrastructure — Memory Card, naming guides, SOPs |
| 10–19 | Administration & Operations | 11 Legal · 12 Finance · 13 Compliance · 14 HR · 16 Operations · 17 Property Dev |
| 20–29 | Marketing & Communications | 21 PR & SEO · 22 Branding · 23 Board Comms |
| 30–39 | Research & Strategy | 31 Market Analysis & AI Tech · 32 M&A · 33 Vision & Strategy |
| 40–49 | Personal & Executive | 41 Health · 42 Travel · 43 Professional Development |
| 50–99 | Unallocated | Reserved for future expansion |
Block 00 is special — it holds the system itself. You'll never need to create items there. Blocks 10–49 are where your department's work lives.
BLUF — Bottom Line Up Front. Throughout this system, all AI outputs use BLUF format: lead with the answer, then the supporting evidence. No fluff, no preamble. You'll see this everywhere.
Anatomy of a Name — tap each segment
Full name: 12.06 [A] - Finance - Korbay Master Sheets
JD Number Builder
Fill in the fields below and watch your JD name build in real time.
Truth Containers
The 3 Filing Cabinet Rules
One drawer, one purpose
Each Truth Container holds documents for exactly one topic, one entity, and one PARA status. Never mix Korbay Finance with Pleck Autos Finance. Never mix [P] project drafts with [A] baseline records.
Label follows the naming system
Every container gets a full JD name before it receives a single document. A container without a name doesn't exist in the system — it's just clutter.
AI reads from this drawer only
When a Gem is attached to a Truth Container, it reads exclusively from those sources. It will not reach outside. This is a feature, not a limitation — it makes the output auditable and trustworthy.
Setting Up a New Container — 8 Steps
Name it first
Use the JD formula from Module 2. Examples: 12.06 [A] - Finance - Korbay Master Sheets or 16.03 [A] - Operations - Pleck Autos MOT Records. No name = no container.
Set to Custom intelligence role
In NotebookLM: click the gear icon → change Default to Custom. This is the single most important configuration step.
Paste the base prompt
Copy this exactly and paste it into the Custom role field. Then replace the last line with your container's specific purpose.
You are a grounded intelligence system operating within the C2O Group's JD System. You ONLY answer from the sources uploaded to this notebook. You NEVER use general knowledge or training data. Rules: 1. Every claim must cite the specific source document and section. 2. If the sources don't contain the answer, say: "Not found in uploaded sources." 3. Use BLUF (Bottom Line Up Front) format for all responses. 4. Never use fluff, hedging language, or vague qualifiers. 5. If sources conflict, present both positions with citations. Do not silently resolve conflicts. 6. Tone: Professional, concise, data-driven. Your role for this specific notebook: [DESCRIBE THE NOTEBOOK'S PURPOSE HERE]
Role examples by department: Finance: "You are a financial data analyst specialising in the financial records and targets in this notebook." | Operations: "You are an operations analyst focused on trade licences, compliance records, and procedures." | Legal: "You are a legal document specialist focused on contracts, regulatory filings, and compliance evidence." | Automotive (Pleck Autos): "You are a vehicle fleet analyst focused on MOT records, service schedules, and parts inventory in this notebook."
Set response length
| PARA Tag | Response Length | Why |
|---|---|---|
| [A] Stable baselines | Default | Reference lookups — concise answers |
| [P] Active projects | Longer | Need depth and analysis |
| [R] Research | Longer | Detailed exploration required |
| Quick fact checks | Shorter | Fast facts only |
Add audience translator (if needed)
Add this to the base prompt if outputs go to non-technical people, external clients, or regulators.
When responding, avoid academic jargon, technical acronyms, and industry-specific terminology. Explain concepts as if speaking to an intelligent person with no background in this field. Use analogies where helpful.
Skip this if outputs stay internal with your team.
Add sources
Three methods: Upload from device — drag files in directly. Add from Google Drive — connect your Drive for live sync. Discover Sources — let NotebookLM suggest sources for a new research topic.
Validate sources — 4 Power Prompts
Run these in your new notebook to verify source quality before relying on it for real work.
1. Source Coverage Check
Confirms your sources actually cover the topic fully.
"Review all uploaded sources. Do they fully cover [topic]? Identify missing angles. Rate: COMPREHENSIVE / ADEQUATE / GAPS EXIST."
2. Research Road Map Generator
Organises sources into a logical reading sequence.
"Organise the uploaded sources into a logical reading sequence with one-line descriptions."
3. Authority & Credibility Filter
Rates source reliability before you trust the outputs.
"For each source, evaluate who authored it, their expertise, and whether it's primary/secondary/opinion. Rate: HIGH / MODERATE / LOW / UNKNOWN."
4. Contradiction Finder
Surfaces conflicts before they corrupt AI outputs.
"Find disagreements and conflicting claims across all uploaded sources. List each with exact citations."
Finance, Legal, Compliance — run all 4. Routine [R] research containers — Prompts 1 and 4 are sufficient.
Register in Memory Card
Add the new container to the Tool Map under your JD block. If it's a [P] project container, also add it to the Project Status section. The Memory Card only maps what you register.
The AI Workers
How Gems Are Named — tap each segment
Gems have a technical code name. Here's what it means:
Meet the Team — click a card to expand
00.01 [A] - System - [GEM] - The Master Index - System Navigation33.01 [A] - Strategy - [GEM] - The Synthesis Engine - Strategic Analysis12.01 [A] - Finance - [GEM] - The Evaluator - Performance Analysis21.01 [A] - PR & SEO - [GEM] - The Messaging Architect - Strategic Communication13.01 FE_Compliance_X__PROD - [GEM] - The Forensic Auditor - Critical Analysis"A Gem without a Truth Container is just a chatbot. It holds zero facts of its own. Every single Gem in this system reads exclusively from what you attach."
The 3-Step Process
Your Daily Workflow
Four Habits — When Each SOP Applies
#project-name48-Hour Rule: Every fragment in Google Keep must reach a Truth Container within 48 hours. Keep is for capture, not storage. After 48 hours, stale fragments create information debt.
When you have a real task to execute — analysis, writing, review — follow this in order without skipping steps.
Pick the right Gem
Evaluator for targets vs actuals. Messaging Architect for communications. Forensic Auditor for compliance. Synthesis Engine for cross-departmental view. Not sure? Ask the Master Index first.
Attach the correct Truth Container(s)
Check the Memory Card for the right container JD number. Most tasks need one [A] baseline + one [P] current container. The Gem will tell you if it can't find data — that's your signal to check the container.
Issue the command
Use BLUF-style prompts: state what you want first, then context. "Provide a RAG status for Q2 Korbay revenue vs annual target." or "List all Pleck Autos vehicles with overdue service intervals. Context: [any relevant detail]."
Do this at the end of every AI session that produced new information. Takes under 2 minutes.
Run the consolidation prompt
Say this at the end of every session that produced new information. Paste the output into your Memory Card.
Consolidate the new information from this session and provide a text block for me to update my Memory Card.
Update the Memory Card
Paste the consolidation block into the right section: Tool Map, Project Status, or Sync Log. Or use the Sync Tool sidebar if available. The Memory Card is only as useful as the last time it was updated.
Click "Sync with Drive" in NotebookLM
Open every notebook that uses the Memory Card as a source. Click Sync with Drive. This is the one manual step that cannot be automated yet — without it, the containers are working from a stale map.
Run these whenever new sources are added to a container. The AI is only as good as the sources it reads.
1. Source Coverage Check
Do these sources fully cover the topic? What's missing?
"Review all uploaded sources. Do they fully cover [topic]? Identify missing angles. Rate: COMPREHENSIVE / ADEQUATE / GAPS EXIST."
2. Research Road Map Generator
Organise sources into a logical reading order.
"Organise the uploaded sources into a logical reading sequence with one-line descriptions."
3. Authority & Credibility Filter
Rate each source: HIGH / MODERATE / LOW / UNKNOWN.
"For each source, evaluate who authored it, their expertise, and whether it's primary/secondary/opinion. Rate: HIGH / MODERATE / LOW / UNKNOWN."
4. Contradiction Finder
Find disagreements between sources before they corrupt outputs.
"Find disagreements and conflicting claims across all uploaded sources. List each with exact citations."
When to run all 4: Finance, Legal, and Compliance containers — mandatory every time new sources are added.
Routine [R] research containers: Prompts 1 (Coverage) and 4 (Contradictions) are sufficient.
Prove It
Final Assessment
Answer all 10 questions, then click Submit. Unlike the module quizzes, you won't see feedback until the end. Pass mark: 8 out of 10.
Generate Your Certificate
Certificate of Completion
and is qualified to operate the C2O Group's
knowledge management and AI system.