AI Control Center: see what every agent can do, and what it did.
Give your team one place to manage how independently Vertex AI agents can work. Set each agent's trust level, review the work that needs human input, and keep a clear record of agent activity and governance decisions.
As agents do more, oversight has to keep pace.
As AI agents take on more work across Vertex applications, tax teams need visibility into what agents are allowed to do, what they’re doing now, and what they’ve done before. AI Control Center brings trust levels, work requiring review, activity, and history into one governance view.
See how governance works
One place to govern your AI agents.
Manage autonomy, review the work that needs human judgment, monitor agent activity, and preserve a record of governance decisions.
Manage agent autonomy.
Set each agent’s trust level within the levels it supports and adjust it as your team’s confidence and needs evolve.
Review work that needs attention.
See recommendations and other work items that require human input, then accept or dismiss them from the Control Center.
Understand agent activity.
Open an agent session to see the original request, the steps the agent took, and the outcome.
Maintain governance history.
Keep a record of trust-level changes, including who made the change and when.
See what needs attention now.
The Overview tab gives your team a current view across the agents you use, including open work items, active agents, recent usage, and autonomy levels. Filter by status or type to focus on what needs action and see which agent raised each item and when.
Adjust agent autonomy while maintaining oversight.
Different agents support different levels of independence. AI Control Center lets you manage each agent at the level of autonomy and oversight that fits your organization’s needs.
As trust levels change, oversight remains consistent. You can review agent activity, monitor work items, and maintain a record of governance decisions in one place.
Human-directed.
The agent provides recommendations and insights, but a user determines what happens next.
Human-approved.
The agent can prepare actions for review, with users approving or dismissing the recommended next step.
Human-supervised.
The agent can complete supported actions independently, with ongoing visibility and governance.
Understand how work was completed.
When questions arise, you need more than the final result. AI Control Center provides visibility into agent activity so teams can understand how work was performed and what decisions were made along the way.
Open an agent session to review the original request, the steps performed, and the outcome associated with each step. This gives teams additional context for review, governance, and operational oversight.