Canon
The governance layer that makes autonomy safe.
Without governance, AI produces activity without intelligence. Canon is the versioned, machine-readable set of rules that define what AI is authorised to do — and what it must never do. It's the constitution of an operating model.
Why Governance Matters
Constraint enables autonomy.
An agent with Canon can operate independently because its boundaries are explicit. An agent without Canon must be supervised constantly, defeating the purpose of automation.
Governance is not bureaucracy.
Canon is not a policy document nobody reads. It's code-adjacent: protocols, rules, and constraints that agents query before every action. Machine-readable, version-controlled, and enforced.
Institutional memory compounds.
Every closed loop creates memory. Every protocol captures a lesson learned. Canon is the mechanism that turns operational experience into permanent organisational intelligence.
What Canon Contains
Four categories of institutional intelligence.
Canon is not a flat document store. It is structured by category, each with a distinct purpose and enforcement mechanism.
Protocols
Step-by-step procedures for specific operations. When to read them, what they enforce.
Governance Rules
Hard constraints. Violating these is always wrong. The non-negotiable boundaries of the system.
Architecture Docs
System design decisions and their rationale. Read before data, infrastructure, or integration work.
Anti-Patterns
Named failure modes with detection criteria. If you see these patterns, stop and course-correct.
The Knowledge Architecture
Three layers of organisational knowledge.
Canon sits at the centre of a three-layer knowledge architecture. Events are captured, truth is codified, and intelligence is delivered through a unified retrieval layer.
LAYER 1
Event Memory
What was said?
Meetings, emails, conversations — captured automatically, enriched with AI, and searchable. The raw material of organisational knowledge.
LAYER 2
Canon
What is true?
Rules, decisions, protocols, and commitments — the governing source of truth your team and agents operate from. Versioned and authoritative.
LAYER 3
Intelligence
What should we do?
Agents query across both layers, weighting results by authority and recency, surfacing the most relevant answer to any operational question.
An agent with Canon can operate independently. An agent without Canon must be supervised constantly — defeating the purpose of automation.