3 critical needs for agentic AI in the public sector

Investment and adoption of agentic AI represent a critical next phase in the evolution of digital government.

When AI agents are designed and empowered to help citizens with simple yet time-consuming tasks—reporting a broken streetlight, renewing a license, applying for food stamps—they can enhance the efficiency of government services while also relieving some of the burden on overextended employees. 

“Agents will be especially effective at the city and county level,” says Murali Swaminathan, CTO at Freshworks. “They can help you retrieve information or set up appointments without having to interact with a person, reducing a lot of the friction of dealing with the government.”

These are still early days in the shift toward agentic AI, and public agencies typically trail private companies in technology adoption. But governments are beginning to embrace both internal-facing agents that help employees do their jobs more efficiently, and external agents that allow constituents to serve themselves.

Lean teams and resources

At every level of government, the demand for services vastly outstrips the ability of agencies to supply them. AI agents can reduce the personnel needed to serve constituents by automating common processes, notes Pankaj Khurana, vice president of technology and consulting at Rocket, an AI-based recruiting platform.

"One of the biggest pain points in government is the heavy burden of paperwork, especially in processes like benefits applications, licensing, and procurement," Khurana says. “Agentic AI can help by automating repetitive intake and triage tasks. Imagine a system that reads documents, validates compliance, flags missing fields, and routes cases dynamically. That alone can cut processing times in half."

And while large federal agencies like the Centers for Disease Control and the Veterans Administration have begun to use AI to read records and route internal communications, we're more likely to see agentic AI deployed first at city, county, and state levels, says Jeff Le, managing principal at 100 Mile Strategies, a public sector consultancy.  

"You often hear the phrase 'States are the laboratories of democracy,’ but in this context, they're also the pilots and proof of concepts of agentic AI," says Le, who served as a deputy cabinet secretary for emerging technology under former California governor Jerry Brown. "That's because it's simply easier to do at the state and local level."

The datasets at a county permit agency or a registrar's office are typically much smaller and less subject to legal restrictions, says Le, which makes agentic solutions easier to implement. And unlike the federal government, most states have privacy laws that help agencies establish clear guardrails for how sensitive data can and can't be used. 

Eventually, AI agents will help understaffed and budget-constrained public sector organizations support constituents in ways they currently cannot, says Le. For example, autonomous agents at the DMV could expedite the process for obtaining a new driver's license after yours has been stolen and you need to board a flight, he says.

"In an era of massive retirements and reductions in force in the public sector, these tools can help streamline services, especially if it's after official work hours or during emergency situations." 

Read also: Retail customer experience is getting an agentic upgrade

Citizens tangled in bureaucratic complexity

When it comes to customer satisfaction, state and federal government services rank at the bottom of the list, below airlines, utilities, and pay TV providers. The unfortunate result is that many citizens are unable to access the services and benefits they're legally entitled to. 

Agentic AI promises some relief from the bureaucratic quagmire that engulfs people trying to access government services, says Roy Kaufmann, president of the Service Members Civic Relief Act Centralized Verification Service, which allows landlords, creditors, and other private parties to quickly verify a person's active-duty military status.

“Governments can and should take advantage of AI agents to cut down on inefficiencies that slow down everything from processing claims to managing internal workflows,” he says. "Right now, some agencies are experimenting with AI in fairly limited ways, such as chatbots for public-facing services or using machine learning to sort through large sets of documents. These are decent starts, but the bigger opportunity lies in streamlining compliance checks, expediting administrative reviews, and making internal systems more responsive.”

For example, agents can streamline the process of determining whether a person is eligible for certain benefits or programs. By cross-referencing databases and flagging any inconsistencies, AI agents can also rapidly verify someone's identity without unnecessarily exposing that person's sensitive information, Kaufmann adds. 

Perhaps the most visible pain point is citizen service delays, where agencies struggle to promptly answer inquiries or process applications, says Dev Nag, CEO of QueryPal, an AI-powered support deflection platform. Natural language chatbots can handle routine questions 24/7, dramatically cutting wait times and improving the citizen experience. 

"We'll also see more proactive citizen services, where AI agents don't just respond to questions but anticipate needs based on life events, automatically triggering reminders about license renewals or benefit eligibility," Nag adds. 

Balancing autonomy with accountability 

The biggest problems may be deciding which government services can safely be automated, and how to protect the privacy and security of citizens' most sensitive data. 

"The challenge to date is that the federal and state governments have not completely provided guidance on function, best practices, and safeguards," Le says. "The question is, how can governments balance consumer protection but also expand service delivery for constituents?"

Organizations will need to find the right balance between efficiency, transparency, and accountability, Kaufmann says. 

Otherwise, they'll end up creating more problems than they solve.

For example, the decision about whether someone is eligible for benefits must ultimately rest with a human, he warns. “Handing off that responsibility to a machine, no matter how accurate or fast it may seem, is a shortcut that puts the government at serious legal risk."

But for basic services—setting up appointments at the DMV, requesting pothole repairs on your street, answering questions about local services—agents are likely to be where citizens go first, adds Freshworks’ Swaminathan. 

“Governments are highly risk-averse,” he adds. “Nobody wants to get blamed for making a big mistake. But for low-risk tasks, public sector agents should work very well.”