Governed Intelligence
Building the foundational layers for AI in regulated industries
AI capability is growing exponentially. The intelligence infrastructure underneath - how organisations know what their AI is doing, where it acts, and whether the knowledge it acts on is still true - is barely being built. I work on that problem.
Read the latest essay → Open the Knowledge Layer deck → All research →
Essays · A series on governed intelligence
The architectural problem of admissibility in supervised AI - the gap between what models can do and what institutions can defend after the fact, and the substrate that closes it.
Latest · May 2026 · the architecture
Spec Is a View
For AI that acts under supervision, the spec is a view, not the source of institutional reliance. Under it sits a governed claim graph the institution can defend at the moment of the act.
The SDD interface does not change. The source of institutional reliance moves to where governance can attach to it.
Earlier in the series · May 2026 · the constraint
Admissibility
For a defined class of regulated AI work, the binding constraint on deployment is no longer the model. It is admissibility - whether, before the act, the institution can treat the model's output as a permissible basis for action, and stand behind that choice afterwards.
The architecture, as a deck
The Knowledge Layer in IT development
An eleven-slide interactive walkthrough of the governed intelligence architecture applied to software delivery - where coding agents are about to act, and the basis they act on is the layer nobody governs.
The case, the requirements, the gap against the tools an engineering organisation already runs, and the practical build - what a team ingests, builds, and runs to govern delegated AI coding.
Open the Knowledge Layer deck →
Opens full-screen in a new tab.
The thesis
Governed intelligence is the architecture that sits between raw AI capability and the environments where failure has legal, financial, or human consequences.
Not governance bolted on after the fact. Governance as a structural property of how the intelligence is built.
In practice: claims with provenance and decay, decision ownership models, evidence chains a regulator can use, and a circuit breaker that halts agent action when the supporting knowledge is no longer fit for the consequence at stake.
Entry point
MRI — Machine-Readable Intelligence
Most large enterprises - particularly those running regulated operations on legacy systems - lack structural visibility into their own operational reality. Business rules embedded in code nobody reads. Dependencies no architecture diagram captures. Regulatory constraints encoded by people who left a decade ago.
The MRI extracts that tacit and embedded knowledge into a governed graph - claims with provenance, confidence, and lifecycle - and surfaces it through dashboards, reports, and queries. Like a medical MRI, it makes the invisible visible without invasive surgery.
It is the first time the organisation's own intelligence is machine-readable. A bounded engagement that delivers standalone value and preserves every architectural option for the full journey that follows.
The journey
Four stages from visibility to initiative
Each stage delivers standalone value. Each stage architecturally enables the next. Some organisations stay at Stage 0 and use visibility for risk and planning. Others carry the governed substrate forward into modernisation, agent operations, and ultimately governed initiative.
Stage 0 — MRI
Operational Visibility
AI-assisted extraction of structured claims from existing systems - code, documents, procedures, operational records - into a governed knowledge graph consumed by humans. No agents, no autonomous decisions. The organisation sees its own structure for the first time. Backward engineering.
Stage 1
Accelerated Modernisation
The governed substrate now informs the modernisation build. Claims about business rules guide forward engineering. The graph moves from read-only diagnostic to active reference. Faster, more accurate, fewer late-discovered rules forcing rework.
Stage 2
Governed Intelligence
Agents become Players alongside humans, consuming governed Context from the knowledge graph and taking actions with real consequences. The full governance apparatus comes online: epistemic tiers, consequence classes, the circuit breaker, override governance, trace-based feedback.
Stage 3
Agentic Operations
The shift from autonomy to initiative. Agent Players begin identifying what to pursue - surfacing opportunities, anticipating exceptions, proactively challenging stale claims. The substrate compounds. The system is self-funding and self-improving. Intelligence as an emergent property.
Published research
Five papers, five manifestos
A research programme spanning architectural diagnosis, theoretical foundations, practitioner methodology, and the governance lifecycle. Co-authored with Arnaud Gelas. Open access on SSRN.
Paper A
Dynamics Blindness: When AI Is Locally Correct and Globally Non-Compliant
Diagnoses the architectural failure mechanism in enterprise AI - LLMs process tokens without tracing causal chains through organisational dependencies.
Paper B
The Predictive Organization: Architecture for Enterprise Intelligence
The architectural resolution - a tripartite structure coupling neural perception with symbolic reasoning, operating on claims-based knowledge.
Paper C
Build the Medium: Why Organizational Intelligence Is Mechanism, Not Metaphor
Ten independent theoretical traditions converge on the same architectural requirements for organisational intelligence.
The programme
Define the standard. Build the architecture. Certify the people.
Financial services, biotech, aviation - industries where the complexity of the governance problem matches the complexity of the solution. The methodology is published and open. The architecture is designed for regulated environments where compliance isn't a cost centre - it's the licence to operate.
About →