The Knowledge Intelligence Capability

From content management to governed knowledge - a structural shift.

Knowledge Intelligence spans both the explicit and tacit dimensions of organisational knowledge: governed, confidence-weighted, and designed to support decisions that depend on the full range of what an organisation knows.

Any practical path toward this capability has to start with what is most legible: the explicit layer. Content Intelligence is the discipline of extracting meaning from an organisation's content estate at scale, assessing signal strength, evaluating reliability, and producing insight that can justify decisions rather than merely inform them. It is not the whole answer, but it is the necessary first layer.

The existing stack asks
Purpose
How should content be organised and found?
Reliability
Is it accessible and compliant?
Action
Has it been reviewed?
Quality
Is it complete and correctly formatted?
Evolution
Is it up to date?
Content Intelligence asks
Purpose
What does this content collectively indicate?
Reliability
How confident should we be in it?
Action
When does it justify action, and when does it counsel restraint?
Quality
Does it strengthen or weaken decision confidence?
Evolution
Is the concept it represents still valid?

Content Intelligence is a pipeline, not a feature. Raw content enters, passes through meaning extraction and confidence weighting, and emerges as a governed insight - or is withheld when confidence is insufficient. Every output carries a weight. The system governs what each confidence level is permitted to influence.

Threshold
Signal strength
Governance action
0.95 +
High Confidence
95%
Auto-validated
Trusted to inform decisions without human review
0.60 - 0.94
Moderate Confidence
72%
Human review
Flagged for expert validation before influencing decisions
Below 0.60
Low Confidence
32%
Withheld
Logged but not permitted to influence any decision
Confidence thresholds are illustrative of the framework architecture, not empirically derived benchmarks. Organisations set thresholds relative to their domain risk profile.

Bringing tacit knowledge into the framework

The most difficult part of Knowledge Intelligence is tacit knowledge: what people know but have not written down. This cannot be fully captured. It can be managed through a metadata-first approach that does not require blanket reading of private content - reducing tacit risk without eliminating the need for human judgement.

01
Digital Signals
Use metadata already produced by work systems: authorship, edit history, usage, and query patterns.
02
Inferred Expertise Map
Map who knows what by combining contribution history, interaction patterns, and domain-level confidence outcomes.
03
Targeted Elicitation
Use short interviews only where risk is high or signals are weak. This is exception work, not the default method.
04
Exit Governance
Treat knowledge transfer as a required control at transitions, with accountable ownership and completion checks.

The Knowledge Health Score

Every knowledge asset within a KI framework produces a single composite indicator - the Knowledge Health Score - combining five dimensions into one signal that tells the organisation whether to trust, review, or retire what it knows.

Composite indicator
Knowledge Health Score

A governance signal that continuously combines five dimensions, then classifies each asset into an explicit operating state: trust, review, or retire.

Confidence
How well-evidenced and validated the knowledge is, distinguishing justified understanding from assumption.
Currency
How recently the knowledge was validated relative to the domain's observed rate of change.
Usage
Whether the knowledge is actively informing decisions, or sitting unused in the estate.
Concentration
Whether critical knowledge is held by too few people or repositories, creating governance risk.
Alignment
Whether the knowledge remains relevant to current priorities, strategy, and operating context.
Output signal
TrustDecision-safe with routine monitoring
ReviewUse with validation and owner attention
RetireDo not influence decisions without remediation
The state is continuously recalculated as evidence changes, so degraded assets stop silently influencing high-stakes decisions.
CONFIDENCE CURRENCY USAGE CONCENTRATION ALIGNMENT KHS
Illustrative profile
Confidence0.86
Currency0.76
Usage0.82
Concentration0.61
Alignment0.71
Radar shape makes asymmetry visible instantly. A single weak dimension can force review even when aggregate health appears acceptable.

The Living Taxonomy

Every knowledge management system ever built has the same structural failure: a taxonomy that someone designed, published, and then watched decay. As the organisation changed, the classification didn't. Knowledge accumulated in categories that no longer fit, under concepts that had been superseded, in structures built for a version of the business that no longer existed. Maintaining it required manual effort that no one had time for. So it was left. And the estate quietly became ungovernable again.

A Knowledge Intelligence framework breaks this cycle through a fundamentally different model. Classification is not designed once and imposed on the estate - it emerges continuously from the estate itself. The intelligence engines that process incoming knowledge read patterns across the full estate, identify where concepts are converging or diverging, detect when new themes are forming before they have been named, and update the taxonomy accordingly. The result is a classification system that evolves in step with the organisation - not months or years behind it.

Humans set the parameters, review proposed changes, and retain authority over the taxonomy's direction. But the intelligence layer does the continuous work of observation and proposal. The taxonomy becomes a living model - not a published artefact that someone has to remember to update.