Data Governance Is Not Agent Governance

I spent years as an information/data architect, so data governance has been part of my life for a long time.

Lately I keep seeing "data governance" and "agent governance" used as if they're the same thing. They're not. The distinction is the whole point of why I'm writing this.

Data governance is about the asset. Who can access it, whether it's accurate, where it came from. Agent governance is about the action — something with permission to reach into systems and act, often with no human in the loop.

The line that stuck with me: you can know exactly what data an agent touched and still have no idea what it did with it. One discipline protects a table that sits still. The other protects something that acts.

Data governance is mostly settled. Agent governance is wide open, and a lot of teams haven't figured out who owns it.

For architects and data folks, here's the encouraging part: the skills transfer. Lineage, access control, classification. All of it. The thing you're governing just changed shape.

Here is the distinction side by side:

Data governance — about the asset

• Who can access it
• Whether it's accurate
• Where it came from

Agent governance — about the action

• What it is allowed to do
• Whether a human is in the loop
• What it actually did

Curious how others are drawing this line — extension of data governance, or its own discipline?

First shared on LinkedIn, where I post regularly on data engineering, AI, and governance.