The SECI Model: Why Your Organisation's Knowledge Spiral Needs a Graph

The Knowledge That Walks Out the Door

Every time an experienced engineer leaves your company, a knowledge graph that was never written down walks out with them. The debugging intuition, the organisational history, the tacit understanding of why-that-component-was-designed-that-way — it vanishes. This is not an HR problem. It is a knowledge architecture problem.

Ikujiro Nonaka and Hirotaka Takeuchi first formalised this tension in 1995 with their landmark work The Knowledge-Creating Company. At its heart is the SECI model, a framework that describes how organisations convert tacit knowledge (personal, context-specific, hard to articulate) into explicit knowledge (codified, transferable, auditable), and back again, in an ever-expanding spiral. Three decades later, the model is more relevant than ever — especially in the age of knowledge graphs, generative AI, and GraphRAG.

This article walks through the four SECI modes, explains why most companies stall at Externalization, and makes the case that a knowledge graph is the closest thing to a general-purpose SECI infrastructure.

The Knowledge Staircase: From Data to Competence

Before diving into the model, it helps to understand where knowledge sits in the broader information hierarchy. Klaus North's knowledge staircase, referenced widely in the German KM literature, frames it succinctly:

Layer	Description	Example
Data	Raw, unprocessed symbols	`42`, `"Berlin"`, `30984`
Information	Data with context and meaning	`"Patient 30984 is located in Berlin"`
Knowledge	Information connected by experience and expectation	`"Berlin has the highest density of radiology clinics, so Patient 30984 has shorter travel times"`
Competence	Knowledge applied to achieve a goal	A sales lead is prioritised based on the Berlin cluster insight
Competitive advantage	Competence that differentiates	Faster prioritisation than competitors

Knowledge, in this view, is information networked through experience. It is relational by nature — and that is exactly where graph thinking enters.

Tacit vs Explicit: The Two Faces of Knowledge

The SECI model rests on a fundamental distinction borrowed from Michael Polanyi (1966):

Explicit knowledge is codified. It lives in documents, databases, manuals, formulas, and code. You can search it, copy it, audit it, and transmit it without loss. "We know what we know, and we wrote it down."
Tacit knowledge is personal. It lives in mental models, intuitions, physical skills, and cultural know-how. Polanyi's famous aphorism — "we know more than we can tell" — captures its essence. You cannot Google someone's debugging intuition.

Most organisations are excellent at managing explicit knowledge. They have wikis, document repositories, and — increasingly — knowledge graphs. But the 80% of valuable knowledge that is tacit remains locked in individual heads, invisible to the organisation.

The Four Modes of Knowledge Conversion

The SECI acronym stands for Socialization, Externalization, Combination, Internalization. These are not sequential stages in a linear process; they are interacting modes in a dynamic spiral. Knowledge moves through them, expanding in quality and quantity as it cycles.

Socialisation (Tacit → Tacit)

Socialisation is the transfer of tacit knowledge through shared experience. It happens when a junior developer pair-programs with a senior, when a surgeon observes a mentor in the operating theatre, or when a community of practice meets for a brown-bag session. No documents change hands. No explicit knowledge is produced. The transfer happens through observation, imitation, and practice.

Key mechanism: Shared experience, not language
Typical failure: Socialisation is time-bound and scale-limited. You can pair-program with one person at a time. Without a recording or documentation layer, the knowledge dies when the expert leaves.
Technology support: Video recordings, collaboration platforms, virtual reality training — but fundamentally, socialisation resists full automation.

Externalization (Tacit → Explicit)

Externalization is the critical bottleneck. It is the process of articulating tacit knowledge into concepts, models, diagrams, or written form. When an engineer writes a runbook for a deployment process they have done a hundred times, that is externalization. When a domain expert sits with a knowledge engineer to define ontology classes and relationships, that is externalization.

This mode is the hardest because it demands that people make the implicit explicit — a cognitively costly act. Nonaka and Takeuchi emphasised dialogue, metaphor, and analogy as tools. Modern approaches add structured knowledge capture: interviews, after-action reviews, and ontology design sessions.

Key mechanism: Articulation through dialogue, metaphor, modelling
Typical failure: People do not externalise what they do not realise they know. Organisations that skip this step never build institutional memory.
Knowledge graph connection: Externalization is where you define your ontology — the node labels, relationship types, and property constraints that turn fuzzy expertise into queryable structure.

Combination (Explicit → Explicit)

Combination is the synthesis of existing explicit knowledge into new, more complex forms. A weekly intelligence report that aggregates data from multiple dashboards. A market analysis that merges patent filings, clinical trial results, and conference presenter lists. A GraphRAG system that runs community detection on a knowledge graph and surfaces emergent patterns.

This is where organisations tend to over-invest because it feels productive — you are building systems, writing reports, and producing artefacts. But combination without prior externalization is rearranging deck chairs.

Key mechanism: Sorting, adding, combining, categorising explicit knowledge
Typical failure: Creating reports and dashboards that compile information nobody externalised in the first place
Technology support: Knowledge graphs, GraphRAG, BI tools, document management systems, LLM-based summarisation

Internalization (Explicit → Tacit)

Internalization closes the spiral. It is the process of embodying explicit knowledge back into personal tacit knowledge. An engineer who reads a well-written runbook, runs the deployment steps, and develops an intuitive feel for the system has internalised that knowledge. Training programmes, simulations, and on-the-job learning are all internalization mechanisms.

This mode is often overlooked. Organisations invest heavily in documenting knowledge (externalization) and building knowledge bases (combination), but they forget that the knowledge must be re-absorbed by people to create competence.

Key mechanism: Learning by doing, reflection, embodiment
Typical failure: Documentation exists but nobody reads it. Knowledge bases grow stale because they are not integrated into workflows.
Technology support: Interactive training modules, LLM-powered Q&A over knowledge bases, automated onboarding agents

The Spiral: Why It Is Not a Circle

The SECI model is often drawn as a four-quadrant diagram, but that misses the point. Knowledge creation is a spiral, not a cycle. Each pass through the four modes elevates the knowledge to a higher ontological level — from individual to group to organisation to inter-organisation.

An individual engineer's tacit insight (socialisation with a peer) becomes a documented runbook (externalization), which gets combined with other runbooks into a deployment handbook (combination), which the team reads and practices until it becomes second nature (internalization). The next spiral starts from a higher base: the team now has shared mental models that enable deeper socialisation.

Dimension	Spiral thinking	Cycle thinking
Knowledge level	Expands and deepens each iteration	Returns to the same state
Scale	Individual → group → organisation	Same scale throughout
Outcome	Growing collective competence	Perpetual documentation churn

Where Most Organisations Stall

In practice, most organisations get stuck at one of two points:

The Externalization Gap. They never capture tacit knowledge systematically. After-action reviews are skipped. Runbooks are out of date. Domain experts are too busy to document. The result is an organisation that is brilliant in its people and mediocre in its systems.
The Internalization Desert. They document everything but nobody consumes it. The knowledge base becomes a graveyard of PDFs. New hires learn through oral tradition rather than reading the hard-won documentation left by their predecessors.

Why Knowledge Graphs Are the Missing Infrastructure

A knowledge graph supports all four SECI modes in a way that traditional document management cannot:

Socialisation is made observable. Graph-based activity logs and expertise location queries (e.g. "who has contributed to the most nodes in the Radiology Ontology?") make tacit expertise findable.
Externalization is structured. Instead of dumping prose into a wiki, domain experts define nodes, labels, and relationships — a form of externalization that is precise and machine-readable.
Combination is automated. Graph algorithms — community detection (Leiden), centrality, path analysis — surface patterns that no human would spot. GraphRAG retrieves subgraphs deterministically, eliminating the hallucination risk of vector-only RAG.
Internalization is queryable. A knowledge graph answers questions directly (MATCH (p:Process {name: "Deploy"})-[*1..3]->(t:Tool) RETURN t) rather than forcing users to read through pages of documentation.

Concretely, a Neo4j knowledge graph stores the explicit knowledge (nodes, properties) and the relationships that encode the tacit understanding of domain experts (edges with semantic types). When a new engineer queries the graph, they benefit not just from the facts but from the structure — the connections that embody how an expert thinks about the domain.

The Modern SECI: A Graph-Enabled Spiral

If we map the SECI spiral onto a graph-enabled organisation, it looks like this:

Socialisation generates new tacit knowledge in teams.
Externalization captures that knowledge as ontology definitions, knowledge graph nodes, and relationship types.
Combination runs community detection and GraphRAG queries across the graph, producing structured insights that no single person could assemble.
Internalization surfaces those insights through graph-powered Q&A systems (LLM + Cypher), embedded in daily workflows.

Each iteration expands the graph — more nodes, richer relationships, higher-quality connections. The graph becomes not just a repository but the organisation's externalised collective intelligence.

Summary

Mode	Conversion	Key Challenge	Graph Role
Socialisation	Tacit → Tacit	Scale and capture	Expertise location, interaction traces
Externalization	Tacit → Explicit	Articulation effort	Ontology modelling, node-edge definition
Combination	Explicit → Explicit	Synthesis at scale	Graph algorithms, GraphRAG, pattern detection
Internalization	Explicit → Tacit	Consumption and practice	Queryable Q&A, embedded retrieval

The SECI model is three decades old, but its diagnosis has never been more acute. The knowledge that differentiates your organisation is mostly tacit, mostly invisible, and mostly at risk. A knowledge graph gives you the infrastructure to capture it, connect it, and spiral it upward — before the next experienced engineer walks out the door.