Special Briefing: AI in Government—More Efficiency but Fewer Workers?

The launch of ChatGPT in 2022 sparked an “AI arms race” with trillions of dollars in investment and profound implications for productivity and employment. For a discussion of how generative AI and related technologies are reshaping public sector operations, workforce needs, and infrastructure planning, Penn IUR and the Volcker Alliance convened a panel of government, academic, and industry experts for “Special Briefing on AI in Government—More Efficiency but Fewer Workers” on November 20, 2025.

William Glasgall, Penn IUR Fellow and Public Finance Adviser at the Volcker Alliance, and Susan Wachter, Co-Director of Penn IUR, co-hosted the Special Briefing. The panel included Jon Hartley, Policy Fellow, Hoover Institution; Leigh Palmer, Vice President, Google Public Sector LLC; Megan Kilgore, City Auditor, City of Columbus, Ohio; Howard Neukrug, Executive Director, The Water Center at Penn and Professor of Practice, Department of Earth and Environmental Science, University of Pennsylvania; and Thomas Sanchez, Professor, Landscape Architecture and Urban Planning, Texas A&M University.

Hartley framed the discussion by presenting new data on AI adoption and implications for the labor market. “As of the end of September, about 37 percent of American workers claim to be using generative AI at work,” he said. Adoption surged early in 2025 but plateaued over the summer, particularly among large firms. According to Hartley, there hasn’t yet been a substantial impact to pay, job openings, or the total number of jobs. He cited recent research showing only “small negative effects concentrated among young people.” Workers report using AI to complement tasks. Nevertheless, “It's still a TBD in terms of what the overall labor market impact is going to be… so far, I think we’re just in the early innings of what’s going to be a much longer baseball game,” said Hartley.

Palmer shifted the conversation to the public sector with a discussion on how local governments can effectively prepare for AI. “Everybody thinks about AI in terms of the models, but you need to prepare for a full stack” when getting ready for AI, she explained. “The bottom layer of the stack is the infrastructure and the compute>power that you're going to need in order to power the tools that you are deploying to your environment.” 

Then comes data and analytics, “where the magic is going to start to happen. So many public agencies have their data locked away in silos,” said Palmer, identifying a key challenge and opportunity of AI deployment. Next in the stack are models like Google’s Gemini, platforms for access, and finally “the pointy end of the spear, where people are really interacting with these tools—the agents and applications where the interface with a citizen may happen.” Palmer highlighted real-world applications: The Chicago Transit Authority built a chatbot to provide timely train and bus information. In Illinois, the Department of Human Services now includes on-the-fly language translation. In New York, AI is being utilized to enhance cyber defense, enabling agencies to respond more quickly to digital threats. “We're just at the tip of the iceberg on the potential for this technology,” said Palmer.

“When you want to know something about what a state or city is really doing, you turn to the auditor or the comptroller, because they have their eyes and ears everywhere,” said Glasgall, setting the stage for Kilgore’s remarks on the transformative impact of AI on government operations in Columbus. “As City Auditor, I oversee the entire revenue side for one of the fastest-growing cities in America,” she said. As the leader of a “large, very financially- and technology-forward office, we've embraced modernization,” Kilgore said, stressing the need for a new mindset: “We need more goats—not sheep. People who are curious, courageous, and willing to climb into unfamiliar terrain as technology reshapes how we’re constantly working and ultimately serving the public.” 

She also called for a shift in budgeting priorities: “We need to view building human infrastructure as well as investing in forward-thinking AI technology as on the same level of vitality as investing in that traditional form of hard infrastructure… Equally as important, governments have to start investing in the skills necessary to allow our public sector leaders and our workers here to really use those tools well: data literacy, ethical reasoning, and creative problem solving. I do believe AI will absolutely redefine public service.” Columbus has built a “digital twin” of its economy using AI to support modeling and projections, she said.

Neukrug introduced the newly launched Water AI Nexus Center of Excellence, focused on driving “equitable, groundbreaking solutions for both water for AI—minimizing water use in AI data centers—and AI for water—leveraging AI to tackle critical water challenges.” Data centers “generate an awful lot of heat, and water cooling is preferred over air cooling because it uses less energy and is less expensive. But this strains local supplies, especially in drought-prone regions… Our goal is to achieve net-zero or even net positive water using all different types of technologies,” said Neukrug. “Our goal is to work with local communities to ensure that data centers fully explore all available strategies to achieve or at least move forward toward net-zero water use.” For water, AI can optimize water systems using real-time sensor data, predict demand, and enable autonomous operations such as smart irrigation and wastewater treatment. “AI-powered digital twins can simulate entire water networks, helping planners test technologies virtually before real-world deployment, saving time and resources,” Neukrug said. 

Sanchez spoke on AI’s dual role in urban planning. “We’re using sensors, we’re monitoring activity and then capturing information and in some cases using it in real time… distinguished from how we are using that information to plan for cities, or to look to the future.” Like Kilgore and Neukrug, he pointed to the potential of digital twins as powerful planning tools to simulate “what-if” scenarios. “What if we build these big data centers in a particular part of a metropolitan area? What does that do to water supply, energy rates, employment, environmental, social, and economic impacts?” Generative AI is reshaping planners’ day-to-day work now. “Planners generate a lot of text. We do a lot of communication. We think about different scenarios. We’re seeing that [AI] tools help us draw on large databases and knowledge across lots of different cities—not just our own.” 

Sanchez explained the critical role for AI in the process of “building scenarios, communicating them to our stakeholders, getting feedback, and getting that back into the planning process for decision-making.” Sanchez echoed other panelists’ uncertainty around future trends and workforce impacts: “We’re still at the beginning of a lot of this, and we don’t know what the overall impacts are going to be, but we know they’re going to be fairly significant in terms of workforce.”

Wachter introduced the open discussion segment of the panel with a question from the audience: “When AI automates administrative functions, what metrics should governments and do governments use to determine whether the administrative function is actually improved?” Hartley emphasized the time-saving potential of AI innovations: “Conditional on using generative AI to complete a task, roughly two-thirds of the time that would traditionally be dedicated to that task is saved,” with the largest productivity gains for lower-income workers.

Glasgall posed a question on the role of education and training in AI adaptation: “How do we get people ready for this new job market?” Palmer pointed to free AI certification programs developed in collaboration with the Virginia state government as a pioneering example. Kilgore shared two on-the-job training efforts adopted in Columbus, including “Auditor Lab, a combination of training at basic levels all the way up to advanced for different technologies, programs, and apps,” as well as a “concerted effort to train critical thinking, helping people ask good questions.” 

This Special Briefing was the 63rd episode in a series of 60-minute online discussions featuring distinguished guests from Penn IUR and Volcker Alliance’s national research networks, along with other leading academics, economists, and federal, state, and local leaders. These convenings are made possible by funding from The Travelers Institute, members of the Penn IUR Advisory Board, and the Volcker Alliance.

Recordings of the entire Special Briefings series are available on the Volcker Alliance or Penn IUR websites. Be sure to subscribe to the Special Briefing podcast, available on Apple Podcasts, Spotify, TuneIn, iHeart Radio, and more.

Subscribe to the Penn IUR newsletter or Volcker Alliance newsletter for advance notice of new Special Briefings.

Related Materials

Making AI Work for the Public: An ALT Perspective

How AI Is Transforming the Way Government Works

The Labor Market Effects of Generative Artificial Intelligence

Water-AI Nexus Center of Excellence

Planetizen, “Four Reasons Urban Planners Can’t Ignore AI”

Tom Sanchez, Artificial Intelligence for Urban Planning (Routledge, 2025)

Route Fifty, "California rolls out AI-powered ID verification for benefits"

Governor Glenn Youngkin Unveils New “Virginia Has Jobs” AI Career Launch Pad in Partnership with Google