aXentic Labs · Blog

Notes on governing an AI workforce

What the governance category inherited from model risk, why it sits in the wrong place for agents, and what it takes to run one in production.

TopicsAgentic AI 1AI governance 1Enterprise AI 1RSS
Coverage map of an agent run. Before deployment (model card, fairness review, sign-off) and after the run (audit log, drift monitoring, incident review) are marked COVERED. The run itself — calls a tool, moves funds, updates a record, notifies a customer — is marked NO CONTROL SURFACE.

· 3 min read · Anantha Padmanabham

The governance we bought was built for models, not agents

AI governance grew up around model risk — evaluate the artifact, document the training data, watch for drift. Agents don't produce predictions. They take actions, and the stack never moved with them.

Read it →