How Federal Leaders See AI as a Catalyst for their Mission

Federal IT officials are iteratively improving artificial intelligence as they leverage the tech to augment citizen and employee services, leaders said Thursday at GovCIO Media & Research’s AI in Action Workshop in Washington, D.C.

“At IRS, we are consciously learning so we’re not trying to be the person without the gate. We’re learning from people who’ve already done things, growing and testing, the things that we want to do in [AI implementation],” IRS Executive Director of the Office of Online Services Karen Howard said at the event.

Smarter Self-Service and Burden Reduction at IRS

IRS is adopting AI methodically and cautiously, according to Howard. She said that the service is focusing on practical use cases to directly support taxpayers and employees. Howard explained that over 160 billion individuals interact with the IRS annually, primarily for four “jobs to be done:” filing taxes, getting refunds, paying money or obtaining information. She added that the overwhelming majority of requests fall into those four categories, and the IRS AI strategy boosts taxpayer self-service in pursuit of those jobs.

“You can pump any of your actions with the IRS into those four things,” said Howard. “In understanding that and boiling it down to a vision for how we leverage AI … it’s really centered around how we help people accomplish those four things.”

Howard said her agency has deployed AI at key taxpayer touchpoints, including voice bots for the Advanced Child Tax Credit and Automated Collection System to help reduce administrative time spent for taxpayers. IRS has also used the tech to identify underreported income, she added.

“Ultimately, our goal is to use AI — generative and predictive — to reduce burden: burden on the taxpayer and burden on our employees, not to add complexity,” said Howard.

As of earlier this year, Howard said IRS has 101 active AI use cases, with 57 more in development. Howard underscored that the IRS is using a “measure, test and learn approach” that recognizes the need for governance and human oversight in AI systems.

“We’ve learned that things like data governance, transparency, explainability, humans in the loop are critical and crucial, and have to be part of our overall AI inaction strategy,” Howard said. “We’ve also learned that hallucinations are real, biases are real, and what AI provides back is only as accurate as the people behind it.”

DISA Maximizing Decisions for the Warfighter

The Defense Information Systems Agency (DISA) is enabling AI to augment the agency’s mission of delivering IT services and defensive networks to the Defense Department. According to DISA CTO and Director of Emerging Technology Steve Wallace, AI is “maximizing decision space” for warfighters and ensuring commanders receive timely, trustworthy information to make superior decisions and gain a strategic advantage.

“The faster we get that information to [the commander] and the more trustworthy that information is, the better decision that they can make,” Wallace said. “The best case is when the user doesn’t even know that [AI is] there.”

Wallace noted that AI within cyber defenses is certainly not a new concept. Predictive and inferential AI models are embedded in sensors, at the network perimeter and in more complex tools within the DOD ecosystem, using AI integration in remote browser isolation (RBI).

“We use these RBI techniques to create some separation in the user’s browsers. Obviously, that’s where a lot of threats make their way into the environment,” said Wallace. “One that we’ve tested and has actually worked really well was using AI, using computer vision, to look at a page, see what’s actually going on the page.”

Wallace added that this real-time assessment allows for dynamic access control — denying access and moving beyond static categorization.

“[It might] change the interaction between the user and the site, making the site maybe view-only or even stripping elements away from the site in real time,” he said.

DISA is also leveraging AI for internal support. Chatbots are helping to answer common employee questions like, “How do I rent a government car?” or “How do I change my work schedule?” The AI integration automates answers to internal human capital questions, freeing up employees and allowing them to focus on mission-critical activities, he said.

Wallace stressed that DISA views AI as a way to replace “toil with action,” increasing responsiveness across the department.

“[These are] basic HR-type questions that exist in a policy somewhere, within a repository somewhere, but these are where we’re actively using chatbots now and getting better about it, to bring that information to the forefront,” said Wallace.

Scaling AI in Government

American Institute of Artificial Intelligence CEO Ali Naqvi advised agencies to focus on research and education in AI to to compete with adversaries like China and Russia.

“Research is very important. You will get lot of soft opinions about AI, but when it comes to real competition, it’s adversaries like China and Russia,” Naqvi said. “At the end of the day, nothing matters except … tech capabilities, your mathematics capabilities and quantum capabilities.”

According to Josh Phifer, principal solutions architect and lead for civilian agencies at Elastic, AI should be used to “bridge the skills gap,” helping individuals become better and understand how to leverage technology more effectively.

“It also helps you understand how you implement more things within your organization, how to actually go about scaling your own infrastructure as well as your own internal knowledge,” Pfifer said. “There is maybe a little bit of a problem with some people outsourcing their critical thinking to AI. We definitely don’t want to do that.”

Glenn Parham, former DOD technical lead for generative AI and current CEO of GovBench, urged leaders to be inquisitive about the workforce’s capabilities in AI.

“Go find and figure out in these CIO shops who are going to be the champions for AI,” said Parham. “Who is actually going to help you get this across the finish line, and who are going to be the early adopters?”

Agencies need to fundamentally shift workforce infrastructure to scale AI. Howard said that organizations need to follow two initial steps: recognizing organizational readiness and identifying skill gaps.

“Once you get employees on board to help you understand where you need to start and how to continue to increase the knowledge and the training to support the tools, you’ll be able to scale rapidly and with logic and solve real world problems,” Howard said.

DISA approaches scaling with a “small but mighty team” that understands AI applications but also recognizes the full challenges faced by various departments. He said that his team partners with “problem owners” who understand their operational issues but lack the technological know-how.

“People are super excited. They want to do this. You have a workforce that is uber engaged,” said Wallace. “Let’s take advantage of it and let’s figure out how to make the most of it.”