How Agentic AI Will Transform Local Government: A Forewarning

Cities and counties have a big opportunity to capitalize on artificial intelligence, if they do so carefully.

Twitter

Alan R. Shark is an Associate Professor at the Schar School of Policy and Government for George Mason University.

Once again, state and local governments across the country are under pressure to do more with less. Predicted budget shortfalls, rising service demands and chronic staffing shortages create a perfect storm — and in that storm, artificial intelligence is often seen as a lifeboat. But most of today’s AI solutions are tools that wait for human prompts. What if AI could go further? What if it could act as a teammate, proactively helping agencies accomplish their goals? That’s the promise of agentic AI — a new generation of intelligent systems designed not just to assist but to act.

Five years ago, before the advent of generative AI, a California county’s chief innovation officer identified a problem and a solution using AI. The problem was that the county (like most) had most of its records siloed off from one another. If someone were picked up on the street by law enforcement, their records would be limited to the self-contained database of that agency. Officers would not be aware of an individual’s mental health, employment, housing, social services or counseling history.

Click the banner below for insight into using AI for citizen services.

Using AI, the county was able to cut across every record that a citizen might be part of, thus providing a more comprehensive picture of what the person may be going through. The goal was to provide a holistic approach to understand their residents better and therefore be better able to provide more targeted and meaningful remedies. With generative AI, that same county could summarize information for the various agencies serving the public and keep everyone informed.

By adding agentic AI, an individual could go online to inquire about certain benefits and programs and find themselves being asked questions to lead them to the right place. Going a step further, the agentic AI might suggest programs and services that the individual was unaware of or thought they were ineligible for. The system could ask, “Would you like me to help you fill out the application now?” Or it might state, “Are you aware that you are eligible for another program or service?”

This scenario can be accomplished without human intervention at any time of day or night, and in multiple languages if programmed accordingly. Instead of waiting for a prompt, agentic AI anticipates what the user is seeking and automatically guides them through a process, including the actual completion of a form or application. Of course, time-consuming, specialized and more difficult cases would be handled by humans.

AI Agents Are More Than Simply Smart Tools

Agentic AI refers to systems that operate with a degree of autonomy, making decisions and taking actions on behalf of users within defined boundaries. Unlike traditional AI systems that analyze data or answer questions, agentic AI can initiate tasks, monitor for changing conditions and adjust its behavior to achieve desired outcomes.

Think of an intelligent agent that watches city infrastructure data in real time and automatically schedules preventive maintenance for a water pump before it fails. Or a system that monitors grant databases, drafts applications based on city needs and submits them when human reviewers give the green light. These aren’t just “smart tools”; they can be considered digital teammates. This isn’t science fiction. The building blocks of agentic AI already exist and are being tested in both private and public sector contexts. The shift is happening, but slowly and unevenly.

The appeal of agentic AI to government is straightforward: better service delivery with fewer resources. These systems can proactively address needs before they escalate; automate entire workflows (not just individual tasks); and provide consistency in high-volume, rules-based tasks.

DIVE DEEPER: Prepare your IT infrastructure for AI.

For example, instead of simply flagging building permits that are overdue, an agentic system might automatically reach out to applicants for missing information, conduct basic compliance checks and issue approvals once all conditions are met. That’s not just automation, that’s operations management at digital speed. These agents can operate 24/7, reducing the backlog of citizen requests; improving response times; and freeing up government employees to focus on tasks that require human judgment, empathy or creativity.

Done right, agentic AI enhances, not replaces, human roles. It can become a force multiplier for strained public teams.

Done wrong, agentic AI systems can have the opposite effect, eroding public trust, providing false and misleading information, and diminishing employee morale.

Click the banner below to sign up for the StateTech newsletter for weekly updates.

Agentic AI Creates Opportunities and Pitfalls for Governments

For every stated benefit, there is a countervailing risk or drawback. And this is where the forewarning comes in. Defining boundaries as to what agentic AI systems have access to and are expected to perform requires a high level of educated oversight and constant monitoring.

Imagine a case where a generative AI system over-promises on a particular benefit, misinterprets an eligibility requirement or enrolls residents into programs for which there is no funding.

Lack of understanding and expertise. Many governments lack the technical expertise to evaluate, manage or oversee agentic AI tools effectively. Equally concerning is the overreliance on vendors, which can lead to black-box solutions with limited transparency or accountability.
Citizen trust and accountability. Citizens may be uncomfortable or skeptical about machines making decisions, especially in sensitive areas such as law enforcement, social services or housing. Further, when AI systems act autonomously, it’s harder to trace and explain decisions, raising accountability and liability concerns. Today, many agentic systems are not designed with sufficient logging or documentation for effective oversight or redress.
Cybersecurity and safety risks. Agentic AI that acts independently may lack sufficient human override or emergency-stop mechanisms. This raises questions regarding “failsafe” mechanisms. Cyber experts worry that autonomous systems can be manipulated, spoofed or misdirected if not appropriately secured.
Data quality and integration barriers. With agentic AI, it becomes even more critical to acknowledge the phenomenon of “garbage in, garbage out.” Poor-quality or siloed data can undermine the effectiveness of agentic AI systems. Finally, tech leaders need to navigate the potential (and likelihood) of interoperability issues. Integrating AI agents with legacy systems can be costly and technically challenging.

DISCOVER: Counties are leading innovation in public services with AI.

Agentic AI is coming, whether local governments are ready or not. The question is not if public agencies will use autonomous systems but how they will do so responsibly and effectively. Cities and counties that take the lead in piloting these systems — starting small, focusing on transparency and integrating human judgment — will not only improve operations but also rebuild trust in government itself.

To be clear, agentic AI is not designed to replace humans altogether. It’s designed to help govern better, more responsively, more equitably and more intelligently.

Register for a Public Technology Institute webinar on agentic AI led by Alan Shark on Nov. 5.

wildpixel/Getty Images