By PYMNTS | October 19, 2026
Highlights
Virginia used agentic AI to scan regulatory documents, cut permit processing times by nearly 80%, and uncover more than $1.4 billion in annual savings, and Boston’s CIO is using an Anthropic agent to analyze city data and inform policymaking in real time.
A WEF framework mapping 70 government functions found half are ready for agentic AI deployment now, with appointment management, document validation, and public information provision leading the way.
More than half of federal agencies are planning agentic AI pilots but fewer than a third have oversight frameworks in place and fewer than a third have a documented kill switch procedure.
Government services have always moved slowly. Long queues, paper forms and manual reviews define an experience most citizens dread. Tighter budgets and shrinking workforces are making that model harder to sustain. AI agents are arriving in the public sector not as a convenience but as a structural response to a workforce that can no longer keep up with demand. The question is no longer whether governments should deploy agents. It is how to do it in a way that is fast enough to matter and careful enough to hold up.
A World Economic Forum article describes the first systematic framework for assessing where governments should start with agentic AI. The report maps 70 core government functions against two criteria: the potential for agentic AI to add public value and the complexity of deploying it responsibly.
The article says half of the 70 functions fall into high or medium readiness. Appointment management, document validation and public information provision come out particularly well. These are high-volume, rule-based functions where agents can make a visible difference quickly and without excessive risk. The article notes that 90% of the 350 public-sector organizations surveyed plan to explore or deploy agentic AI within two to three years, while Gartner predicts over 40% of projects could be cancelled by 2027, often because organizations moved without a clear sense of where the value lies.
Manuel Kilian, managing director for the Global Government Technology Centre in Berlin, told the WEF that agents allow governments to move from automating individual tasks to delivering entire outcomes. The article’s practical path: assess local conditions honestly, develop risk strategies before deployment, sequence from high-readiness functions first and test through small pilots before scaling.
States Are Already Running Agents on Public Services
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
A Route Fifty article covers a new Urban Institute playbook for state and local agentic AI adoption and profiles three live deployments. In Virginia, a pilot program used agentic AI to scan regulatory documents, uncovering more than $1.4 billion in annual savings and reducing permit and license processing times by nearly 80%. In Utah, an autonomous system handles routine prescription renewals in partnership with health platform Doctronic, with physicians agreeing with 91% of the AI’s decisions. In Boston, CIO Santiago Garces integrated agentic AI into the city’s open data portal using an Anthropic model context protocol server to analyze city data and inform policymaking.
Graham MacDonald, chief information officer at the Urban Institute, told Route Fifty that the governance challenge is real: human reviewers who do not understand how an agent works cannot meaningfully oversee what it produces. The playbook recommends agencies document each tool’s intended purpose, target population, measurable outcomes and estimated costs, and establish seven core roles to ensure agents continue operating adequately over time.
Federal Agencies Have the Appetite but Not the Kill Switch
A Nextgov article reports that 53% of federal IT executives surveyed are exploring or actively planning agentic AI pilots. AI use across federal agencies more than doubled from 2024, with agencies reporting more than 3,000 use cases.
But only 20% have defined pre-deployment testing policies. Only 8% have an incident response framework. Fewer than a third have documented kill switch procedures. Mike Hurt, Global Vice President of U.S. Public Sector at ServiceNow, told the publication that 77% of federal leaders say oversight frameworks are essential, yet fewer than a third have implemented them.
The data readiness gap runs deeper. ServiceNow Federal CTO Jon Alboum said the answer is bringing data and workflows together in a single environment so policies can be applied and controls enforced across the organization. Built-in accountability, the survey concludes, is the prerequisite for any agency serious about deploying agentic AI at mission speed.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.