Federal agencies have spent the past two years using generative AI tools like Microsoft 365 Copilot and chatbots summarize information and answer questions. Now, agencies are starting to explore agentic AI, systems that go beyond assistance and begin taking actions across systems based on defined rules. As this shift begins, the focus is no longer just what AI can create, but what actions should be allowed to perform.

• Examined how AI agents move from insight to action across Government systems
• Understood the oversight measures required when software makes decisions, including authority, oversight and rollback
• Identified where agentic AI delivers the most value
• Learned what it takes to maintain trust, auditability and mission assurance as AI becomes more autonomous