Foundations: The Thinking Behind the Framework

Knowledge Strata didn’t invent the classification of organisational knowledge. It operationalised it for the AI era.

Theoretical foundation

Knowledge Strata draws on a long lineage. Drucker (1959) named the structural distinction between knowledge work and manual work. Polanyi (1966) added the foundational observation: “we know more than we can tell”. Nonaka & Takeuchi (1995) gave the field the SECI model: how tacit and explicit knowledge convert. Collins (2010) split the tacit half into three operationally distinct types: Relational (unsaid but tellable), Somatic (compiled into the body), and Collective (residing in group practice).

Knowledge Strata adds an operational layer. The four-layer classification — L1 (transferable), L2 (domain), L3a (documented organisational), and L3b (tacit organisational) — integrates these foundations into a single working vocabulary. Each boundary has a diagnostic test, designed to apply to specific work. Each layer has a role in deciding where AI fits and where it doesn’t.

On boundaries: the edges of classification

Every classification framework produces boundary cases. Good frameworks acknowledge this as a feature, not a flaw.

Cynefin (Snowden, 1999) builds ambiguity into its model as a fifth “Confusion” domain, legitimising “I can’t yet tell which domain this sits in” as a valid first-pass answer. Knowledge Strata treats boundary cases the same way. When a piece of knowledge resists clean classification, the boundary itself tells you something useful: it’s at the edge where one kind of expertise transitions into another.

In practice: two people disagreeing about whether something is L1 or L2 will, in resolving it, reveal the specific judgement that differentiates generic work from specialist work in their organisation. That revealing is the point.

The thinkers behind the lens

Five thinkers whose work forms the irreducible foundation of Knowledge Strata. Each gets one line on what they contributed and one line on where it anchors in the framework. A broader further-reading list — contemporary convergent thinkers, practitioner frameworks, and adjacent academic work — is maintained in the methodology document for readers and practitioners who want to go deeper.

Michael Polanyi

1966

The Tacit Dimension

“We know more than we can tell.” The foundational observation that personal knowledge exceeds articulable knowledge. The philosophical starting point for the whole tradition Knowledge Strata rests on.

Anchors:Anchors the existence of L3b: the knowledge that can’t be fully written down.

Ikujiro Nonaka & Hirotaka Takeuchi

1995

The Knowledge-Creating Company

The SECI model — Socialisation, Externalisation, Combination, Internalisation — describing how knowledge converts between tacit and explicit forms.

Anchors:The earliest of the three generations Knowledge Strata rests on. SECI gave the field the general vocabulary for how tacit and explicit knowledge move. Collins (2010) later built on it by refining the tacit half into three operationally distinct types.

SECI describes knowledge as a four-stage cycle between tacit and explicit forms. Organisations that execute Pillar 1’s Extract well activate the full loop. Captured knowledge distributes widely, sparks new thinking in the people who work with it, and becomes the next generation of tacit expertise. Knowledge Strata doesn’t replace SECI. It sits inside it.

Harry Collins

2010

Tacit and Explicit Knowledge

Three operationally distinct types of tacit knowledge: Relational (unsaid but tellable), Somatic (compiled into the body), and Collective (residing in group practice).

Anchors:The direct academic source of our Ask / Observe / Develop classification of L3b. Collins built on Nonaka & Takeuchi’s 1995 SECI model, splitting the single ‘tacit’ category into three types with transfer mechanism baked into each. The action-verb labelling (Ask / Observe / Develop), single diagnostic question, and integration with the wider L1 / L2 / L3a / L3b classification are our contribution. The underlying typology is Collins’.

Gary Becker

1964

Human Capital (Nobel 1992)

Formalised the distinction between general training — portable across organisations and industries — and organisation-specific training, valuable only within one organisation.

Anchors:The economic foundation for why L1/L2 (transferable, portable) and L3 (organisational) are structurally different kinds of asset. Knowledge Strata operationalises sixty years of labour economics for AI-era decisions.

Erik Brynjolfsson, Danielle Li & Lindsey R. Raymond

2023

Generative AI at Work (NBER Working Paper 31161)

Field study of 5,179 customer-support agents using generative-AI assistance. Average productivity gain of 14%. But 34% for novices versus minimal gain for experts.

Anchors:The empirical anchor for the L3a / L3b split. AI compressed the Layer 3a learning curve (documented procedures that novices were still absorbing) but did not replicate the Layer 3b judgement that experts had already internalised. Measurable evidence that the codified / tacit boundary is where AI-scale gains begin and end.

A full annotated bibliography is in preparation.