software-architecture.aiYour brand must not live only in your head.
Knowledge management with AI systems for founder-led SMBs: You know why one supplier was kept and another let go. As long as that lives only in your head, your brand is a single point of failure. I build the knowledge management system that makes your decisions retrievable for team and tools. Without freezing them.
Values decisions
Which collaboration to decline · which supplier to keep · how to communicate during a crisis
Context knowledge
Why this wool producer in Portugal · why natural rubber instead of synthetic · why this packaging
Heuristics
How complaints get classified · how trend requests get evaluated · how influencers get filtered
Brand memory
What has been tried · what was community-loved · what didn't work
Three pillars, one system.
Founder knowledge
Knowledge out of your head
Four recurring patterns.
What happens while the brand knowledge stays in your head.
You get pinged privately, even off-hours
On values decisions someone else can't or won't make alone. Influencer requests, unusual collaborations, edge cases. Your reply takes seconds. The fact that it's needed every weekend is the issue.
Onboarding runs as a one-person show
New hires legitimately ask a lot. Why this supplier, why this material, why this packaging. You explain it every time yourself. That doesn't scale, and the cost shows up only when you're already exhausted.
Decisions in the team drift
When someone agrees to a collaboration in your absence that you would never have agreed to, it's rarely tone-deafness. It's missing context. What you implicitly know, the team doesn't.
Service replies lose context
A complaint gets a polite, correct reply. What's missing: that the customer has ordered repeatedly, that something similar came up before. That kind of information often only sits in your head, not in the ticket.
Two options as long as the knowledge stays unstructured: decide everything yourself and never quite step away, or delegate and clean up afterwards. Both cost. One time, the other trust.
What knowledge management for founders actually is. And what it explicitly isn't.
Quick clean-up, because there's a lot of confusion here. Founder knowledge isn't what you know. It's what you do when deciding, without noticing.
- Not a Notion wiki nobody maintains and that turns into a graveyard of outdated pages after three months
- Not a custom GPT that gets the brand manual as context and pretends to be you
- Not an onboarding PDF the new hire reads on day one and forgets by day three
- Not SOPs (standard operating procedures), because a brand doesn't have operating procedures, it has values and heuristics
A documented system of four knowledge layers.
Values decisions
What you held, what you rejected. And why.
Context
The stories behind decisions. What shaped the brand.
Heuristics
The rules of thumb you no longer apply consciously.
Brand memory
Anchors, anecdotes, brand moments. The invisible continuity.
Founder knowledge is less what you know.
More what you do when deciding, without thinking about it.
Four layers. So not all knowledge gets treated the same.
Not every piece of knowledge belongs in the same drawer. Values decisions need a different format than supplier context, and heuristics work differently than brand memory. These four layers have proven to be a robust split.
Values decisions
Which collaboration to decline · which supplier to keep · how to communicate during a crisis
Context knowledge
Why this wool producer in Portugal · why natural rubber instead of synthetic · why this packaging
Heuristics
How complaints get classified · how trend requests get evaluated · how influencers get filtered
Brand memory
What has been tried · what was community-loved · what didn't work
- 01
Values decisions
The decisions in which your brand defines itself. Which influencer yes, which no. Which supplier stays, even if more expensive. How you communicate in a crisis. This layer is the most important, because this is where your brand emerges, and at the same time the hardest to document, because the decisions are often gut decisions. The job of the layer is translating gut into criteria.
- 02
Context knowledge
The reasoning behind your material, supplier and product choices. Who sits in Portugal and why. Why this material and not the other. Which packaging serves which trade-off. This layer is the knowledge base your team can answer from when customers ask, without you in the loop.
- 03
Heuristics
Your micro-decision rules. When someone writes that the product feels different, what are the three questions you ask yourself. When a trend surfaces, what are the two filters you apply before deciding whether to participate. When an influencer reaches out, what are the five points you check within two minutes. These heuristics run on autopilot in you. Documented, they become useable for your team.
- 04
Brand memory
The institutional memory of your brand. Which launch did not go as planned and why. Which campaign was a community favorite, and what were the ingredients. Which step was reversed and what was the lesson. This layer prevents your team from repeating mistakes in two years that you already solved.
How do you capture founder knowledge without cloning you.
This is the trickiest question of the whole project. My explicit aim is not to replicate you. My explicit aim is to mirror how you would have decided yesterday, so your team and your tools can build on that today. What you decide tomorrow stays your call.
AI-supported knowledge interviews
Three to five sessions of sixty minutes each, where you talk and I take notes and probe with the help of a language model. You tell, I structure, the model helps surface gaps. Concretely: I probe ten of your recent values decisions, have you describe how you respond to a specific complaint, and walk through your supplier list. From these sessions emerges the raw material of the four layers.
Slack and email archive analysis
With your approval, I review the last twelve months of Slack and email traffic where you commented on values decisions in a structured way. This is a goldmine, because that is where your gut shows up in real sentences, often more precise than how you would phrase it in an interview. GDPR-compliant, processed locally, with a clear deletion deadline after project end.
Decision diary over three weeks
For three weeks you document every values decision you make in two sentences. What was the question, what did you decide, why. That is effort, honestly about fifteen minutes a day. But it is the only method that captures your gut in real time, before you have rationalized it.
Negative library of misjudgments
As important as anything else. Which three decisions would you make differently today, and why. These negative examples sharpen the system more than ten positive ones. They tell the founder brain what does not belong to your brand, which makes the filter logic realistic.
The system mirrors what you would have decided yesterday. What you decide tomorrow stays yours.
How it works
Six steps. Clear sequence. You see every intermediate state.
- 01Step 1
Knowledge audit
Map the pain density. What are the most frequent questions landing in your Slack today. Which decisions got escalated to you in the past three months that didn't need to. Where did you delegate and correct after the fact. Output is a knowledge heatmap of your current reality.
- 02Step 2
Knowledge interviews
Three to five sessions of sixty minutes, in which the raw material of the four layers gets generated. In parallel, the decision diary runs for three weeks, if you commit to it.
- 03Step 3
Structuring into four layers
Values decisions, context knowledge, heuristics and brand memory get filled in, with concrete examples from your material. This phase is intense, because this is where the translation between gut feel and documentation happens.
- 04Step 4
Engine build
Notion structure, RAG system, Slack integration. Optional: voice bridge to Pillar 1, service bridge to Pillar 2. You see intermediate states, you test, you give feedback.
- 05Step 5
Pilot with one person from the team
One person from your team uses the founder brain in daily work, three to four weeks. They log: which question got answered well, which not, where they still asked you. This phase fine-tunes the system before the whole team gets access.
- 06Step 6
Roll-out and update loop
The team gets access. Quarterly update sessions are built in: new values decisions, new heuristics, new brand memory. Plus a clear escalation logic: when must you decide yourself, when may the system answer.
Knowledge Inventory
By the end you have a map of your implicit knowledge. You see what only lives in your head, what's already partly documented, and which layer needs to be lifted first.
What you get
- 2 × 90 min founder interviews (remote, structured across four layers)
- Layer map: values, context, heuristics, memories — weighted by risk
- Prioritised lever list: which layer to document first
- Structured doc set in Notion, Confluence, Obsidian or the documentation tool of your choice, as a starting point for your own upkeep
Format
2 × 90 min interviews, remote only. 1 week processing.
Fixed price
€1,490
plus VAT, one-time.
Frequently asked
Fair question, because RAG is being worn out as a buzzword. The difference is in the substance and in the architecture. A custom GPT with the brand manual as context sounds hollow after three days. It doesn't know anything real. What I build is more than classic vector retrieval: an information network of values decisions, context knowledge, heuristics and brand memory. Vector search is just one layer. On top of it come knowledge graphs, rule logic and context-sensitive routing. So the system actually knows why your wool producer sits in Portugal. Not just that they're there.
If you recognise the pattern.
A first call is 15 minutes, free and without a sales pitch. We check whether the setup fits, how deep the four layers need to go, and whether to build Pillar 3 before or after the other two. A subsequent knowledge audit typically takes two to three weeks. For a more concrete first read with samples from your current knowledge sources, a 60-minute knowledge diagnosis is available for 99 €.