After twenty years of consulting, I had knowledge scattered everywhere. Client frameworks in old slide decks. Industry insights in email threads. Lessons learned in notebook margins. Workshop methodologies in documents I could not find. My expertise was real, but accessing it required remembering where I put things, which is not a system. It is a prayer.
Building a knowledge base with AI changed this. Not just organizing existing knowledge, but making it queryable. I can now ask my knowledge base questions and get answers drawn from two decades of accumulated expertise. It is like having a research assistant who has read everything I have ever written, attended every meeting I have been in, and organized it all for instant retrieval.
Why a Knowledge Base Is Not Just Filing
Let me distinguish between what most people think of as a knowledge base and what I am describing.
Most knowledge bases are organized filing systems. Documents sorted into folders, tagged by topic, maybe with a search function. You put information in and retrieve it by browsing or searching for keywords. Better than nothing, but limited by the fact that you need to know what you are looking for.
An AI-enhanced knowledge base is something different. You put information in, and then you ask it questions. Not keyword searches. Natural language questions like “What did I learn about pricing strategy from the Vulpine Creations experience?” or “What frameworks have I used for client onboarding across different project types?”
The AI does not just find documents. It synthesizes answers from across your knowledge base, combining insights from multiple sources into a coherent response. It is the difference between a library (you find the book) and a knowledgeable advisor (they give you the answer, drawing from everything they have read).
This distinction matters for founders because your expertise is not neatly organized into topics. A lesson about pricing might come from a magic product launch, a consulting engagement, and a startup accelerator experience simultaneously. An AI knowledge base connects these disparate sources in ways that folder-based organization cannot.
What Goes Into the Knowledge Base
The first step is feeding the system. Here is what I include and how I organized it.
Process documents and SOPs. Every standard operating procedure I have written. These form the operational backbone of the knowledge base. When I ask “How do I onboard a new consulting client?” the answer comes from the actual SOP I follow.
Client work products. Proposals, reports, strategies, and recommendations (anonymized where needed for confidentiality). These contain applied expertise: not theory, but how I actually solve problems for real clients.
Blog posts and articles. Everything I have published. My written content represents my thinking on topics I care about, organized and refined into publishable form. This is some of the most useful content in the base because it is already structured and articulate.
Meeting notes and summaries. Especially from strategic meetings, client sessions, and mentoring conversations. These capture real-time thinking and decisions that formal documents miss.
Personal notes and reflections. Rough notes from books I have read, conferences I have attended, and ideas I have had. These are the raw material that has not been refined into formal content but still contains valuable insights.
Voice and style guides. Detailed documentation of how I write, speak, and present — tone examples, vocabulary preferences, structural patterns, things to avoid. This is what keeps AI-produced content sounding like me rather than like generic AI.
Industry research and references. Data, reports, and analyses relevant to my areas of expertise. These provide the factual foundation that supports my experiential knowledge.
Source materials from book projects. The roughly 3,200 source pieces I compiled for my six books are all in the knowledge base. They represent the most organized version of my accumulated expertise.
Total volume: roughly 2,000 documents across all categories. This sounds like a lot, but it accumulated over twenty years. The AI ingestion process took about a week of focused effort.
The Technical Setup
I want to be practical here because the implementation details matter.
Storage layer: Obsidian, a markdown-based note-taking application. All documents are stored as plain text files in a folder hierarchy. I chose Obsidian because it is local-first (my data stays on my machine), searchable, and works with any text-based AI tool.
Organization: Folders by category (clients, processes, content, research, personal). Tags for cross-cutting themes (pricing, marketing, Austrian-market, AI, leadership). A simple naming convention: YYYY-MM-DD-description.md.
AI interface: When I want to query the knowledge base, I do one of two things:
Option A (for specific queries): I open the relevant documents and paste them into Claude along with my question. This works for focused queries where I know which documents are relevant.
Option B (for broad queries): I use a local AI tool that can search across all my documents and synthesize answers from multiple sources. As of 2026, tools like Claude Projects, NotebookLM, and custom RAG (Retrieval Augmented Generation) setups handle this well.
Cost: Obsidian is free. The AI tools I use are part of my existing tech stack subscriptions. The incremental cost of maintaining the knowledge base is effectively zero.
For founders just starting, you do not need a sophisticated setup. A folder of text files and a paid AI subscription is enough. You can add sophistication later as your needs grow and the tools improve.
How I Use the Knowledge Base Daily
Let me show you the practical value through actual use cases from a recent week.
Monday: Client preparation. I had a call with a new potential client in the tourism sector. I queried the knowledge base: “What experience do I have with tourism businesses? What frameworks have I used for tourism-related projects?” The AI pulled from three different sources: notes from a Startup Burgenland tourism startup, a consulting project for a hotel group, and research I had compiled on DACH tourism trends. In five minutes, I had a thorough brief on my own relevant experience. Without the knowledge base, I would have relied on memory and missed at least one of those sources.
Tuesday: Content creation. I was writing a blog post about financial planning for startups. I queried: “What specific examples do I have of financial projections that were significantly wrong? What lessons did those teach?” The AI surfaced four examples from different contexts, two of which I had forgotten about. The resulting article was richer and more specific because of those examples.
Wednesday: Proposal writing. A client needed a proposal for a brand strategy project. I queried: “What brand strategy frameworks have I used? Which produced the best results?” The AI compared three frameworks across multiple engagements and recommended the one with the most consistent positive outcomes. The proposal referenced specific past results, which strengthened the pitch.
Thursday: Problem-solving. A client was struggling with team alignment. I queried: “What approaches have I seen or used for team alignment challenges?” The AI pulled from consulting notes, book research, and personal reflections to produce a menu of options with context about when each works best.
Friday: Learning reflection. I had attended a webinar on AI regulation in the EU. I added my notes to the knowledge base and queried: “How does this new regulation information affect my current advice to Austrian startups?” The AI connected the new information to existing knowledge and flagged three areas where my advice might need updating.
Each query took two to five minutes and produced results that would have taken thirty to sixty minutes to compile manually, if I could have compiled them at all. The knowledge base does not just save time. It surfaces connections and context that I would not have remembered or discovered on my own.
Building the Base: The Practical Process
If you are starting from zero, here is the process I recommend.
Week 1: Core documents. Add your most important and most frequently referenced documents. SOPs, client templates, your best writing, and your core frameworks. This gives you a useful knowledge base immediately.
Week 2-3: Historical archive. Go through old files, emails, and documents. Anything that contains expertise or experience, add it. Do not spend time reformatting. Drop documents in as-is. The AI can work with messy formats.
Week 4: Organization. Add tags, categories, and a basic folder structure. This makes targeted queries more effective because you can direct the AI to search specific categories.
Ongoing: Continuous addition. Every meeting summary, every new document, every note from a conversation gets added to the knowledge base. This is the habit that makes the base increasingly valuable over time. I add roughly five to ten documents per week.
When the weekly inflow gets larger — voice memos, email threads, whiteboard photos — I batch it and let AI do the categorization in one weekly pass:
<source_material>
[All raw capture items placed at TOP of prompt — meeting notes,
emails, voice memos, screenshots]
</source_material>
<existing_categories>
[Summary of current knowledge base structure and recent entries]
</existing_categories>
<instructions>
First, quote the most significant passage from each capture item.
Then categorize each item by type: process, client, domain, market, lesson.
Extract key information and identify connections to existing entries.
Flag items that need human review before filing.
</instructions>Two techniques matter here. First, source material goes at the top of the prompt with instructions at the bottom — with fifty capture items, this ordering produces more accurate categorization because the AI reads all the material before it starts processing. Second, the “quote before categorizing” instruction anchors the AI in the actual content of each item rather than letting it assign categories from surface-level pattern matching. My review of the AI’s processing takes about 30 minutes, and it’s where I add the layer the AI can’t know — what the client didn’t say, what I think their real concern is.
Month 2 and beyond: Extract your tacit knowledge. The hardest knowledge to capture is the stuff you do instinctively but have never articulated. The most effective method I’ve found is a structured AI interview, guided by few-shot examples:
<interview_context>
You are extracting domain expertise from [your name], who has [years]
of experience in [field]. The goal is to capture tacit knowledge —
things they do instinctively but haven't articulated.
</interview_context>
<few_shot_examples>
<example>
<question>What's the most common pricing mistake you see in Austrian
startups?</question>
<answer>They price based on their costs plus a margin, not based on
what the customer's problem costs them. A startup solving a EUR 50,000
problem should not be charging EUR 500. I saw this with 31 of the 44
startups I worked with at Startup Burgenland.</answer>
</example>
<example>
<question>When a client says "we need better marketing," what do you
actually hear?</question>
<answer>Usually they mean "we're not getting enough leads" which usually
means "our offer isn't compelling enough." Marketing is rarely the real
problem. The offer is the real problem.</answer>
</example>
</few_shot_examples>
<instructions>
Ask questions that uncover: common mistakes, patterns observed,
contrarian views, frameworks developed from experience, and
decision-making heuristics. Match the depth and specificity of
the examples above.
</instructions>The few-shot examples show the AI the level of specificity you want. Without them, it asks generic questions that produce generic answers. With them, it asks the pointed questions that draw out the experience-based knowledge that makes your expertise valuable — and the structured Q&A output goes straight into the knowledge base.
Monthly: Maintenance. Review recently added documents for accuracy and completeness. Remove outdated information. Update documents that reference things that have changed.
The knowledge base is never “done.” It grows with your experience. The longer you maintain it, the more valuable it becomes, because the AI has more context to draw from when answering your queries.
Common Mistakes
Adding everything without any structure. A knowledge base with 5,000 untagged documents is hard to query effectively. Basic organization (folders and tags) dramatically improves the quality of AI responses.
Capturing knowledge that isn’t yours. If a fact is easily searchable online, don’t replicate it in your knowledge base. Focus on what’s unique to your operation — your processes, your clients, your domain insights, your lessons. Generic knowledge has generic value.
Not updating or removing outdated information. A knowledge base that contains outdated advice alongside current advice will confuse the AI and you. Mark outdated documents clearly or remove them.
Treating it as write-only. A knowledge base you add to but never query provides no value. Build the query habit. Ask at least one question per day.
Over-engineering the setup. Start simple. A folder of text files works. You can add sophisticated tooling later. Shipping an ugly first version applies to knowledge bases too.
Making it solo-dependent. If the knowledge base can only be updated or interpreted by you, it has the same bus-factor problem as knowledge that lives in your head. Design it so another competent person — or a well-configured AI agent — could maintain and use it in your absence.
Takeaways
-
Start building this week with your twenty most important documents. SOPs, frameworks, best work, and core reference materials. A useful knowledge base requires a critical mass of content, and twenty documents is that minimum.
-
Query the knowledge base daily. Before any client call, content creation session, or strategic decision, ask the base what relevant experience or information you have. The habit of querying is what turns documents into accessible intelligence.
-
Add five to ten documents per week from ongoing work. Meeting notes, new insights, client learnings, and any new content you produce. Continuous addition is what makes the base increasingly valuable.
-
Keep the technical setup simple. Markdown files in folders, tagged by category, queried through your existing AI tools. Sophistication can come later.
-
A knowledge base makes your expertise compounding. Every document you add makes every query richer. After six months of consistent use, you will wonder how you worked without it.
