Govern AI agents
AI summary
Maintain visibility into AI agents across your organization by identifying owners and information sources. Policies automatically flag agents lacking ownership, that are stale, unused, public, or relying on public websites and outdated content. Create automations to notify owners and resolve these issues.
Know every AI agent, who owns it and what it answers from. Policies flag agents nobody owns, stale, unused and public agents, and agents that answer from public websites or outdated content.
Who it is for: IT operations, IT security and risk teams, and the teams that own agents.
Steps
- Connect the agent sources. Copilot StudioCopilot StudioCopilot Studio is the Microsoft tool for building your own AI agents. Rencore Governance covers it with the Copilot Studio connector, part of the AI & Agents module.Read more needs Copilot Studio consent and Dataverse access. Claude EnterpriseClaude EnterpriseClaude Enterprise is the business plan of the Claude AI assistant by Anthropic.Read more needs Anthropic consent.
- See what you have. The Copilot Studio inventory lists agents with owner, status and environment. Create a report from a template to share the list.
- Flag the agents that need attention. Turn on the policies below: create a policy from a template.
- Act on them. Create an automation from a template that notifies the owner. For Claude Enterprise, you can delete a Claude project with an automation or from the inventory.
Step by step: Find Agents Nobody Owns and Find Agents That Answer from Public Websites.
Policy templates
- Copilot Studio policies: Orphaned Copilot Agent Ownership Gap, Stale Unpublished Copilot Agents - 30 Days, Unused Agent Cleanup Candidate - 90 Days, Unauthenticated Public Agent Security Risk (see agent authenticationAgent authenticationAgent authentication decides whether users must sign in to chat with a Copilot Studio agent.Read more), External Web Content Dependency Risk, Confidential Site Referenced in Copilot Agent.
- SharePoint policies: SharePoint agent is not approved, SharePoint agent references confidential site without approval, SharePoint agent uses outdated site collection, SharePoint site collection contains more than 10 agents.
- Copilot policies: External Website Data Source Risk.
- Claude policies: Claude project not updated in 90 days, Claude project is empty, Project visible to the whole organization.
- ChatGPT policies for custom GPTsCustom GPTA custom GPT is a version of ChatGPT that someone in your ChatGPT Enterprise workspace configured for one purpose and can share with others.Read more: Stale custom GPT, Custom GPT with an inactive owner, Publicly reachable custom GPTs.
Limits
- Agents built with Agent Builder and agents in Agent 365 are not in the inventory.
- Rencore Governance does not create agents, and it cannot publish or unpublish a Copilot Studio agentCopilot Studio agentA Copilot Studio agent is an AI agent someone built in Microsoft Copilot Studio.Read more.
- Claude Enterprise and ChatGPT EnterpriseChatGPT EnterpriseChatGPT Enterprise is the business plan of the ChatGPT AI assistant by OpenAI.Read more are in Preview. For them, only the objects that their admin APIs return are visible.
Frequently asked questions
Which AI platforms does Rencore Governance support?
Rencore Governance covers Copilot Studio, SharePoint Copilot, Claude Enterprise (in Preview), and ChatGPT Enterprise (in Preview). It provides inventory, policies, and automations for agents across these platforms.
What permissions are needed to connect Copilot Studio?
You need to grant Copilot Studio consent and Dataverse access. For Claude Enterprise, you need to provide Anthropic consent to connect those agents.
What policy templates are available for Copilot Studio?
Templates flag orphaned agents, stale agents, unused agents, public agents, external web content risks, and confidential site references. Create a policy from a template to monitor these issues.
Which agents do not appear in the Rencore inventory?
Agents built with Agent Builder and agents in Agent 365 are not shown in the inventory. You cannot govern these through Rencore Governance.
How do I automatically notify agent owners of policy violations?
Create an automation from a template. When policies are triggered, the automation notifies the owner that their agent needs attention.
Was this page helpful?
Thanks for the feedback.
Thanks. What was wrong?
Next steps
- Prepare your content for Copilot
- Reduce license and AI spend (Copilot Studio consumption)
- Keep every workspace owned