KNOWLEDGE LAB / FRAMEWORK / AI KNOWLEDGE SYSTEMS / IN BUILD
KARI
Knowledge Architecture Readiness Index
Before you make knowledge agentic, ask whether the knowledge is ready.
Organisations are rapidly adding AI to their knowledge systems.
But those systems inherit the architecture of the knowledge underneath them.
AI doesn't remove knowledge debt. It consumes it.
KARI is a framework I've developed for examining that layer.
Content Models · Metadata · SSoT
WHAT IS KARI?
KARI, the Knowledge Architecture Readiness Index, is a proprietary benchmark methodology for assessing how ready a knowledge system is for both human and machine consumption.
It looks beyond whether documentation exists or whether an organisation has deployed an AI interface. Instead, it asks whether the underlying knowledge has the structure, governance and semantic foundations required for reliable retrieval and increasingly agentic use.
Think of it as a maturity model applied specifically to the Knowledge Layer.
Not:
How mature is your AI?
But:
How ready is your knowledge for what you're asking AI to do with it?
WHY READINESS MATTERS
And a Knowledge Graph cannot repair a domain model nobody has defined.
This creates a dependency chain:
KARI examines the first layer before organisations invest further up the stack.
WHAT KARI EXAMINES
KARI focuses on the structural foundations of a knowledge system, across six dimensions.
Illustrative synthetic profile shown for demonstration only, not derived from any real assessment. Employer-specific scores are never published.
STRUCTURE
How consistently is knowledge organised, modularised and represented? Can both humans and machines identify what a piece of knowledge is?
TAXONOMY
Does the organisation have a controlled way of describing concepts, products, audiences and relationships? Or does terminology change depending on who created the content?
GOVERNANCE
Who owns knowledge? What is authoritative? How does knowledge move from creation through review, publication, maintenance and retirement?
CONTENT MODEL
Does the system understand knowledge as structured entities and attributes, or primarily as pages and documents?
SEMANTICS & ONTOLOGY
Are important domain concepts and relationships explicit enough to support machine interpretation?
RETRIEVAL READINESS
Can the correct knowledge be discovered with sufficient context, provenance and authority to support search, RAG and agents?
FROM DIAGNOSIS TO ARCHITECTURE
The score itself isn't the interesting part.
The gaps are.
A readiness assessment should reveal where structural intervention has the greatest leverage. For example:
This is where KARI connects to the rest of my Knowledge Lab.
KARI IN THE KNOWLEDGE ARCHITECTURE STACK
KARI
Where is the knowledge system structurally weak?
Taxonomy · Content Models · Ontology
What does the organisation know, and how is it structured?
Knowledge Graphs
How do those entities and concepts relate?
Search · RAG · GraphRAG
Can the right knowledge and context be recovered?
API · MCP · Agents
Can machines reliably access and use it?
AI Evaluation
Is the resulting behaviour actually grounded and trustworthy?
CURRENT STATUS
KARI is currently an evolving methodology, being piloted against real-world knowledge systems.
The framework, methodology and scoring model are my own intellectual work. Where the methodology is applied within an employer environment, the organisation's content, findings, scores, architecture and implementation details remain private. They will not be published here.
Knowledge Lab will instead document the framework itself and, where useful, demonstrate its application using public or synthetic examples.
WHAT KARI ISN'T
It's part of the analytical toolbox I bring to Knowledge Architecture: a structured way of determining whether the knowledge underneath an AI system is ready for what we're asking that system to become.
THE THESIS
We've spent enormous effort measuring models.
But there's another variable upstream:
What condition was the knowledge in before the model ever saw it?
KARI is my attempt to make that question measurable.