How agentic AI can improve efficiency and reduce costs for federal agencies

Under increased scrutiny to do more with less, government agencies are grappling to streamline operations, reduce their footprint, and slash expenditures while embracing transformative technologies. Enter agentic AI: an emerging technology that goes beyond traditional automation by using advanced reasoning models, intelligent orchestration engines and integrated tools to execute complex tasks autonomously.

Although computerized agents have been around for decades, they’re a new concept for generative AI. “Agents actually date back to the ‘50s. They were largely rule-based,” says Amina Al Sherif, Generative AI Lead for Google Public Sector, in a recent FedScoop podcast underwritten by Deloitte. “We started seeing machine learning entering the agentic definition in the ‘90s, then influenced by deep learning in the 2000s, and now we’re in what we would call our autonomous era with generative AI agents.”

According to Al Sherif, enterprises are mainly in the early stages of agentic AI maturity, primarily focusing on point-to-point human-to-agent interactions rather than the full potential of collaborative fleets of agents managing multiple autonomous workflows.

Ed Van Buren, AI Leader for Government and Public Services at Deloitte, highlights two areas where agencies can benefit from agentic AI. The first is where the government is running external and internal business processes. “All of those processes have some common attributes,” he says. “There are multiple steps where a request could come in. It goes through reviews and validations and verifications, and in many of those steps, humans are still in the loop.” Agentic AI can augment human workers or even automate administrative tasks, freeing up personnel for more complex work.

The second area is critical problem-solving and reasoning. Van Buren explains that many government employees deal with situations requiring data gathering, analysis, research and decision-making. “Agentic AI is giving them the ability to start executing parts of that activity, to bring together that information and do some preliminary reasoning on it, to accelerate their pace to getting to good answers, and then ultimately to getting to good decisions,” he says. Examples include permitting intelligence and benefits administration.

Looking ahead, Al Sherif stresses the urgency for government agencies to use agentic AI. “If government agencies right now are not already putting serious consideration, if not already implementing agents, they’re going to find themselves behind very quickly here in about six months to a year,” she says. She also anticipates a strong focus on security protocols for enterprise-grade agentic systems, fleets of agents, and more advanced and predictive decision-making agents moving forward.

Van Buren foresees a “big wave” around agentic AI adoption, predicting that more than 15% of work decisions will be made autonomously and the market growing from $5 billion to $50 billion over the next 5 years. Knowing that, he advises leaders to understand the technology to manage it effectively, ensuring agents are focused on the right tasks and deployed wisely, given the critical nature of government workloads.

“I think that technology could go all the way pretty quickly, but we’re going to have to go on a journey of change management to get leaders and the workforce comfortable with the arrival of these tools in the organizations,” says Van Buren. “And I just don’t think we should overlook the change-management elements as we go on what will be a transformational journey for government.”

Listen to the entire podcast conversation here. And learn more about how AI agents are reshaping the future of work.

This podcast was produced by Scoop News Group for FedScoop and sponsored by Deloitte.