More autonomy changes what firms need to govern. When an AI agent can retrieve information, use tools, apply logic, and move through multiple steps of a financial workflow, governance cannot stop at the final output. For higher-risk work, firms may also need visibility into the sources accessed, tools called, results or actions generated, and human reviews completed along the way. That starts earlier than the audit trail. It means defining the agent’s role, setting appropriate permissions and triggering conditions, and deciding where professional judgment belongs. @FactSet MCP provides one concrete example of how a boundary can work in practice: it is provisioned at the user level and is entitlement-aware, helping ensure AI requests draw only on authorized tools and content. Our latest blog explores how financial institutions can translate these considerations into controlled agentic workflows: