EXPERTISE · AGENTIC AI

Autonomy is an architectural and governance decision.

Agentic AI refers to systems that can pursue goals through multi-step reasoning and action, often selecting tools, invoking APIs, using memory and interacting with other systems or agents with varying levels of human supervision.

DEFINITION

What does this mean in an enterprise context?

Agentic AI refers to systems that can pursue goals through multi-step reasoning and action, often selecting tools, invoking APIs, using memory and interacting with other systems or agents with varying levels of human supervision.

CONTROL DIMENSIONS

Four lenses for executive review.

Identity

Every agent needs a distinguishable identity, owner and trust context.

Authority

Permissions should be explicit, bounded, contextual and revocable.

Tools

Tool use must be constrained by task, data sensitivity and action impact.

Oversight

Human involvement should scale with consequence, uncertainty and autonomy.

QUESTIONS TO ASK

Useful questions for leadership, risk and technology teams.

What decisions may the agent make independently?

The answer should identify an accountable owner, a defined control expectation and evidence that the control operates.

What systems and data can it access?

The answer should identify an accountable owner, a defined control expectation and evidence that the control operates.

How are delegated permissions constrained and revoked?

The answer should identify an accountable owner, a defined control expectation and evidence that the control operates.

What happens when an agent behaves unexpectedly?

The answer should identify an accountable owner, a defined control expectation and evidence that the control operates.

RESEARCH

Independent analysis on agentic ai & responsible autonomy.

The research library is published separately under /research/ using WordPress.

Explore research →

INDEPENDENT ADVISORY

Complex technology. Clearer risk decisions.

For AI strategy, governance, Agentic AI, security, assurance or cyber-risk requirements, describe the decision you are facing and the context around it.

Discuss an Advisory Requirement ↗