Human actors self-limit through biology. Fatigue, hunger, and social friction create natural pauses — the gaps inside which oversight actually functions. Autonomous AI systems have none of that. If governance wasn’t designed in before deployment, it doesn’t exist.
Generate a regulator-defensible governance profile for one AI agent — in under fifteen minutes.
The structural gaps that appear when autonomous systems replace human actors. Not technology problems — governance design problems. Controls calibrated for human speed and human judgment that do not hold when the operating entity never stops.
Why those gaps persist. Existing frameworks assume biological constraints that AI agents don’t have. That assumption is not incidental — it is structural, embedded in how regulation, compliance, and organisational oversight were designed. Understanding the mechanism is how you address it at the source.
I publish practical, grounded frameworks to address these gaps. Rewriting process definitions. Redesigning controls for agent-speed operations. Defining what “good” looks like before defining what “fast” looks like. Direction based on what is actually working, not theory.
AI Agents in Accountability-Critical Environments
A practitioner-grounded research programme on governance architecture for autonomous AI systems operating in environments where human accountability is legally and operationally required.
The programme addresses a structural gap: existing governance frameworks were designed around human actors and assume the biological constraints those actors carry. When the actor is an autonomous AI agent, those constraints are absent — and the frameworks have no mechanism to compensate.
Working papers and peer-reviewed publications are forthcoming. The first working paper is in preparation.
A six-part series on AI agent governance — covering the gap between deployment capability and governance maturity, machine-readable boundaries, human oversight architecture, and accountability in multi-agent systems. Each edition includes a downloadable framework.
Why most AI agent deployments fail before they start — and what governance maturity actually looks like.
Edition 2: Decision BoundariesThe difference between capability statements and Decision Boundary Contracts — and why it matters at scale.
Edition 3: Human OversightThe Oversight Spectrum — from autonomous to controlled — and how to match intervention to risk.
The chain of responsibility problem in multi-agent systems. Introduces the Accountability Canvas — four named owners, assigned before deployment.
Read →
Edition 5: Audit TrailsNaming owners is one thing. Proving what happened is another. The Handoff Receipt — a structured log that regulators can actually read.
Read →
Edition 6: ImplementationWhere to actually start. Five primitives, six editions, one implementation path — governance shipped, not launched.
Read →
Four recurring themes across everything we publish.
What changes when the actor doesn’t self-limit. How oversight must be redesigned when fatigue, hesitation, and biological friction are removed from the system.
The difference between governance discovered after failure and governance embedded before it. Why retrospective accountability frameworks are insufficient for continuously operating autonomous systems.
How to keep human judgment meaningfully in the loop when the system operates faster, continuously, and at scale than any human oversight mechanism was designed to handle.
The governance problem of autonomous systems is not unique to any one industry. Financial services, healthcare, legal systems, and critical infrastructure face the same structural challenge — with different regulatory vocabularies and the same underlying absence.
Available On