As state and local governments explore agentic artificial intelligence capable of autonomous decision-making, the Urban Institute urges establishing robust governance frameworks before these systems become embedded in public services.
- Agentic AI pilots in Virginia and Utah show potential cost savings and operational benefits.
- Governments need tailored fairness metrics and clear accountability roles for agentic AI use.
- Democratization of AI development raises risks of errors without proper governance.
Market signal
State and local governments worldwide are piloting agentic AI systems that extend beyond generating content to autonomously planning and acting toward goals. Examples include Virginia’s regulatory review pilot that identified over $1.4 billion in annual savings and Utah’s AI-assisted prescription renewal program, with high physician agreement on AI recommendations. Boston's integration of agentic AI into its open-data portal further illustrates expanding adoption in municipal functions.
These initiatives highlight a growing trend where public sector entities seek to leverage AI not just as a tool but as an operational actor capable of reducing costs and streamlining services. This shift signals increased demand for AI solutions that support autonomous decision-making and operational workflows tailored to public administration contexts.
Operator impact
The added autonomy of agentic AI introduces new risks, especially as AI development and deployment extend beyond traditional IT teams to broader government employees. This democratization accelerates innovation but can increase the potential for operational errors, harm, and workflow disruptions. Sensitive government functions, such as benefits administration, require particularly stringent governance to avoid bias and ensure fairness.
The Urban Institute’s agentic AI playbook advises agencies to document key deployment parameters including purpose, target population, and anticipated outcomes, while separately defining accuracy and fairness metrics relevant to context. Establishing distinct accountability roles such as evaluation leads and transparency leads is critical to maintaining audit trails, monitoring harm potential, and fostering stakeholder engagement.
What to watch next
Market participants should monitor how governments formalize governance frameworks around agentic AI and integrate these into existing IT and regulatory structures. Key developments will include refinement of fairness standards tailored to varied constituent groups and the operational scaling of autonomous AI in public-facing applications.
Emerging regulatory guidance or standards shaped by pilot outcomes in states like Virginia and Utah could influence broader adoption patterns and vendor requirements. Operators providing AI tools should prepare for heightened demands around documentation, transparency, and compliance as governments strive to balance innovation with responsible deployment.