{
  "id": 8952637,
  "title": "Urban Institute Urges Guardrails as Governments Experiment With Agentic AI",
  "url": "https://urgent.news/2026/09/21/urban-institute-urges-guardrails-as-governments-experiment-with",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-21T15:44:49.000Z",
  "source": {
    "name": "PYMNTS",
    "slug": "pymnts",
    "url": "https://www.pymnts.com/news/artificial-intelligence/2026/urban-institute-urges-guardrails-as-governments-experiment-with-agentic-ai/"
  },
  "original_language": "en",
  "account": "State and local governments are increasingly testing artificial intelligence systems with the ability to independently take actions, according to guidance from the Urban Institute. The institute's new playbook aims to help officials prepare for and responsibly deploy agentic AI, which goes beyond generating content to plan and act autonomously toward goals. Unlike standard generative AI, agentic systems can make future decisions and take actions based on that planning.\n\nAgentic AI is already being experimented with in government. Virginia's pilot used agentic AI to streamline regulatory documents and found potential annual savings of over $1.4 billion, while Utah's AI policy office is exploring if autonomous AI can assist with prescription renewals. Boston has integrated agentic AI into its open-data portal to help inform policy queries.\n\nThe issue, as highlighted by the Urban Institute, is ensuring proper governance structures are in place before these systems become embedded in public services. Agentic AI can be developed and deployed outside traditional IT departments, potentially leading to errors, workflow disruptions, and harms, particularly when dealing with sensitive functions like benefits administration.\n\nTo address these concerns, the institute's playbook recommends that agencies document key information about the agent, such as its purpose, target population, outcomes, and costs. Accuracy and fairness should also be defined separately, with accuracy considering the completeness, clarity, and reliability of the agent's output under different conditions. Oversight should increase for systems that will be used by constituents, with fairness metrics tailored to specific populations and services.\n\nThe institute also suggests formalizing accountability by identifying seven roles, including an accountable owner, evaluation lead, security lead, transparency lead, responsible AI lead, contributor, and independent reviewer. Their responsibilities could include maintaining audit trails, conducting evaluations, gathering stakeholder feedback, and monitoring potential harms. The institute stresses the importance of bringing diverse individuals into the process to establish common goals, metrics, and standards.",
  "summary": "State and local governments are beginning to experiment with artificial intelligence (AI) systems capable of taking actions on their own. According to new guidance issued by the Urban Institute, making sure governance structures in place before those systems become embedded in public services is an essential first step. The institute’s new agentic AI playbook is […] The post Urban Institute Urges…",
  "key_points": [
    "Urban Institute releases playbook for agentic AI governance in government.",
    "Virginia pilot shows potential $1.4B savings with agentic AI streamlining.",
    "Boston's open-data portal integrates agentic AI for policy query assistance."
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}