Governable Enterprise Agentic AI
Identity, authority and autonomy - with controls where consequences occur.
Working architecture / research in progress
Explore the modelResearch & thinking
Thinking with consequences.
Published research and evolving architecture models for AI-enabled organisations and systems. Written to clarify difficult decisions, with evidence, assumptions and limitations made explicit.
Research & thinking
Enterprise agentic AI
Enterprise Agentic AI Capability Map
A capability landscape for enterprise agentic AI spanning interaction, orchestration, knowledge, integration, governance and assurance.
Working capability model / research in progress
Explore the map
Identity, Authority and Autonomy
A Cobalt Giraffe control model for enterprise agentic AI built around identity, authority and autonomy.
Working conceptual model / research in progress
Explore the control model
Enterprise Agent Control Plane
A working architecture for separating agent execution from identity, authority, autonomy, policy enforcement, observability and evidence.
Working architecture / research in progress
Explore the control planeEditorial position
Quality before cadence.
Thinking begins with a question, then develops through primary-source research, architectural reasoning and adversarial challenge. Publication follows only when the argument can be supported and defended.
AI may assist with discovery, organisation, challenge and editorial development. AI-generated material is not treated as evidence. Published positions and conclusions remain human-authored and accountable.