Not all tacit knowledge is the same.
Before you decide how to make Layer 3b knowledge visible, you need to know which kind it is. One diagnostic question separates the three approaches.
Where this fits
Pillar 1 tells you which knowledge sits at Layer 3b — the unwritten, organisation-specific expertise that can’t be fed to AI because it doesn’t exist as text anywhere. This page tells you what to do about it. Three approaches, one test, different remediation paths.
Knowledge-management researchers have distinguished these three types of tacit knowledge for decades — most cleanly in Harry Collins’ 2010 taxonomy of Relational, Somatic, and Collective types.
The transfer mechanisms underneath our Ask / Observe / Develop verbs are Nonaka & Takeuchi’s (1995) SECI cycle — Socialisation for Develop, Externalisation for Ask and Observe. What’s new isn’t the classification. It’s that AI has made the question urgent.
Each card maps a diagnostic question to a remediation path. Ask the question of your L3b holder; the answer tells you which approach fits.
Ask · Relational
The knowledge IS articulable — it’s tacit for social reasons (nobody asked, mismatched salience, hoarding), not cognitive ones.
How:A structured conversation extracts it. Often in a single session.
Cost:Low. The knowledge is already cognitively available; the bottleneck is that no one has been asked the right way.
AI role:Once extracted, the knowledge becomes L3a and feeds any RAG system.
Observe · Somatic
The knowledge has compiled through years of practice below conscious access. The expert can DO it but can’t DESCRIBE it.
How:Watch them work. Codify iteratively. Test against real cases. Refine over multiple rounds.
Cost:Medium to high. Requires sustained observation and multiple codification attempts.
AI role:Partial. AI can assist the codification pass (pattern-extraction from observations) but the core extraction is human-led.
Develop · Collective
The knowledge lives in the group, not the individual. Relationship trust, cultural intuition, political judgement.
How:Transfer through immersion — apprenticeship, mentoring, shared practice.
Cost:High and continuous. Can’t be compressed. The relationship context IS the knowledge.
AI role:Structurally limited. This knowledge can’t go into a system. It CAN grow in other people.
The diagnostic question determines the remediation path. The extraction cost is different for each; the AI-readiness implications are different for each.
The hollow codification risk
Documentation without reasoning is worse than nothing.
The most common Extract failure mode: compiling L3b into L3a documents without capturing the why. The resulting documents look complete, the organisation assumes the knowledge is now transferable, and the L3b holder leaves. What stays behind is a shell. Codification is only successful when it captures reasoning, not just outcomes.
Extract is where the real work of Pillar 1 happens.
The framework classifies. Extract moves the needle.