# How to structure your AI Master Prompt ## Introduction There are different approaches to create, structure, document and use your [[AI Master Prompt]]. Here I'll focus on two: - Using a single long document: the "default"/straightforward approach - Using many small documents stored in and leveraging a central knowledge base: my recommended approach ## Basic approach: One long document In its most basic form, an AI Master Prompt is a single note or document that contains all of the information. This approach is a good starting point, and is easy to use (you have a single document to share with AI), but it has important drawbacks and limitations: - It's hard to maintain because the document might become very long and become harder to keep up to date - It forces you to duplicate information in multiple places (e.g., your goals will be somewhere AND in your master prompt) - It's harder to connect to other information in your existing system(s) - It's harder to scale because you can't tell AI to only read some parts of the document, it's generally all or nothing. This means you'll hit the context size limits faster ## Recommended approach: Multiple documents stored in a central knowledge base A much power powerful approach consists in *enriching*, and *leveraging* a [[Knowledge Graph (KG)]] (e.g., [[Personal Knowledge Management System (PKMS)]]). With that approach, the idea is to create notes in an existing knowledge base, covering various topics (cfr [[How to create your Personal AI Master Prompt]] and [[How to create your Business AI Master Prompt]]), and to connect those to other notes/pieces of information in the knowledge base (e.g., information about yourself, about your business, your goals, etc). Thus, your AI Master Prompt becomes an integral part of your knowledge base instead of being an isolated document. Here's a visualization of this approach: ![[AI Assistants - visualization.png]] Cfr [[AI Assistant - Overview.canvas|AI Assistant Overview]] The [[Knowledge Graph (KG)]] thus serves as the main source of information for [[Large Language Models (LLMs)]]. When you want to use your [[AI Assistants|AI Assistant]], you'll have to point it to the right note(s). This approach has multiple advantages - There's no need to duplicate information. You can point to existing notes/documents in your knowledge base - It's easier to maintain, since you have many small documents instead of one. Moreover, those documents are stored along with the rest of your information - It's scalable, because you can tell your [[AI Assistants|AI Assistant]] to only load specific parts and only load the relevant context if and when needed (e.g., using [[Prompt Lazy Loading AI Design Pattern (PLL)]]) - You don't have to upload anything to your [[Large Language Models (LLMs)]]. Rather, you just point it to your entry point and let it fetch information itself from your knowledge base Challenges of this approach: - You need to be able to connect your [[Large Language Models (LLMs)]] to your chosen knowledge base. This may require additional tooling (e.g., [[Model Context Protocol (MCP)]] servers) I have explained how I approach this in the following masterclasses: - [[Knowii Community Event - Knowledge Builders - 2025-06-03 - AI Master Prompt]] - [[Knowii Community Event - Knowledge Builders - 2025-06-10 - Model Context Protocol (MCP)]] ### Create an entrypoint If you decide to split up your AI Master Prompt in multiple notes/documents, you need to create an entrypoint. The entrypoint is the note/document that you point your [[Large Language Models (LLMs)|Large Language Model (LLM)]] to. You should think of that entrypoint as the first lines of a computer program. That's where it all starts. That's where you point to the rest of the elements to loads (e.g., your personal or business notes/documents, etc). The entrypoint must include: - An overview of your knowledge base - Explanations about how your knowledge base is architected/structured - Explanations about how to work with your knowledge base (e.g., where to find notes/documents, how to load those, etc) ## Related - [[AI Assistants]] - [[AI Master Prompt]] - [[How to create your Personal AI Master Prompt]] - [[How to create your Business AI Master Prompt]] - [[How to use your AI Master Prompt]]