What Did Agencies Learn About AI in 2025?

We asked three experts to share their thoughts on what agencies had learned about artificial intelligence this year. Here’s what they said.

Keep Focused on Ethical AI

In theory, agencies should consider the ethical implications of any technology they deploy. But the conversations around AI and ethics have become especially intense given the technology’s tremendous capabilities and potential for misuse. Because one of the most prominent concerns is the misuse of data — especially constituent data — data experts are driving many of those conversations, said Amanda Crawford, Texas Chief Information Officer (CIO) and President of the National Association of State CIOs (NASCIO).

“Across the country, we’ve seen an increase in the role of data officers, of chief data officers and privacy officers,” she said. “The emphasis is there on improving those disciplines.”

But better discipline doesn’t necessarily entail more regulations. The Texas Legislature, for example, is avoiding strict regulations but recently directed Crawford’s office to develop an AI code of ethics, she said. Agencies also can rely on guidance, frameworks and standards that NIST and other groups developed.

You Don’t Need to Go It Alone

Government IT leaders often talk about the value of sharing ideas and lessons learned across federal, state and local governments, but usually it’s more of an ideal than a practice. AI could change that.

AI’s rapid adoption and evolution are pushing government officials to look for insights from their counterparts across the public sector, said Michelle Amante, Senior Vice President of Government Programs at the Partnership for Public Service.

Although their contexts and requirements might be different, they can discuss how they’re using the technology, the constraints they build around it and their strategies for responsible AI. “Sometimes, you don’t even realize that AI can be used to solve a problem until you hear that someone else has used it,” Amante said.

Value Your Employees’ Domain Expertise

When Pennsylvania conducted a yearlong pilot of ChatGPT, state officials learned a lot about its capabilities and limitations. But they also realized just how much using AI successfully hinged on their employees’ expertise.

For example, as part of the pilot, the HR team explored various ways they could integrate GenAI into their workflows to improve the quality of services they offered other agencies, said  Harrison MacRae, Director of Emerging Technologies for Pennsylvania. “They felt these tools could be really helpful in that first 80% of the work, but then relied on their [own] expertise in the [HR] space to make final decisions, to do final checks and to make sure it aligned with what their end users needed,” MacRae said. “This is something we’ve tried to model through a lot of the use cases at the commonwealth,” he added.

AI and the Art of the Possible

We asked GovLoop readers, “What do you see as the most valuable potential benefits of using AI in
government? Select up to three.” Here’s what they told us were their top use cases.

• 57% – Employee efficiency or productivity
• 55% – Operational efficiency
• 26% – Data sharing and collaboration
• 26% – Data-driven decision-making
• 20% – Responsive constituent services

This article appeared in slightly different form in our guide, “Expert Answers to Your AI Questions.” To read more about what leaders are saying about AI, download the guide:

Photo by Campaign Creators on Unsplash