Your AI agents need badges
A new hire started at your company on Monday. No badge. No manager. No record of what it opened or what it changed.
You'd call security. But this one's an AI agent, and today it has more access to your data than half your team.
Nobody put that on the rollout slide. Companies are pointing agents at their data faster than they can govern what those agents do with it.
The gap is measurable
New research this month from BARC: 79% of enterprises say they can scale AI without breaking governance. Only 29% actually know where their data lives. You can't govern what you can't find. And you can't audit an agent you never gave an identity.
So the job changes
Data governance used to mean governing the data. Now it means governing the data and every agent that touches it. Same three questions you'd ask any new hire on day one:
Who is it? A verified identity, not a shared service account.
What can it do? Permissions scoped to the task, not the whole warehouse.
What did it do? A full audit trail. Every read, every write.
This stopped being theory this week
Snowflake shipped Agent Identity at its Summit: verified identity, scoped permissions, and an audit trail for every agent. The catalog is becoming the badge office.
This is the steward's next job
Here's the part most data leaders are missing. The person who used to write the access policy for people now designs the one the agents run under. You don't lose the role. You move up a level. The steward becomes the orchestrator.
The agents are already in the building. The only question left is whether they're badged.
Would yours pass a security review tomorrow?
What's the first control you'd put on them? Tell me. I read everything.