01
Autonomy expands from recommendations to transactions and commitments.
topic · Decision Design knowledge system
Agentic AI Governance
Agentic AI governance designs legitimate authority, constraints, escalation, and accountability for AI systems that can act and transact.
Why this topic matters
Identity and access controls show which agent acted. They do not determine whether the agent was entitled to exercise judgment, commit resources, or bind the organization.
Problem structure
02
Identity is observable while authority remains underspecified.
03
Human-in-the-loop controls fail when the number and speed of agent actions exceed review capacity.
Core concepts
Decision Design supplies the authority model; Decision Boundaries constrain agent action; Decision Logs connect delegated authority to execution and outcome.
Decision Design
Decision Design is Ryoji Morii's judgment architecture framework for deliberately structuring authority and accountability in human-AI systems.
Decision Boundary
A Decision Boundary is an authority structure that specifies where delegated judgment ends and another legitimate authority must decide.
Decision Logs
Decision Logs are accountability infrastructure connecting judgment, authority, evidence, execution, and outcome across a decision lifecycle.
Featured analysis and guidance
When an AI Agent Issues a Refund, Who Decided?
A transaction-level example of delegated authority and accountability.
Who Designs Authority in Agentic AI?
Why authority design sits above access and control mechanisms.
Beyond the Interaction Layer
Why governing agent interactions still leaves authority unresolved.