AI is Not ‘One Size Fits All’ for Challenges, Federal Officials Say

There is no skeleton key to unlock AI’s potential for government, IT leaders from agencies ranging from Veterans Affairs Department to the Air Force Research Laboratory said this week. GovCIO Media & Research’s AI FedLab in Reston, Virginia on Wednesday, executives across government and industry discussed AI applications, pain points and ways to integrate the new technology into their processes.

Over the course of the day, participants said there is still much to learn about AI and its potential. Many of those working with AI share the same challenges, but are exploring different ways of approaching and working to solve them.

Following the discussions, government leads from each roundtable presented shared takeaways from their tables’ conversations with the room.

Addressing Risk and Bias

AI risks depend on context and agencies must carefully consider mission objectives.

“AI may not be the right solution and it usually has a lot of start-up costs, so it’s important to think about the mission that you’re trying to achieve,” said Defense Department Chief Digital and Artificial Intelligence Office (CDAO) Director of Assurance and Assessment (A2) Jonathan Elliot. “AI is usually your last choice to achieve that solution because it has significant upfront investment costs, so look toward what your specific needs are, before implementing.”

Generative AI, while convenient, can create biases or incorrect responses.

“If you’re relying on the data that’s coming out, you have to be cautious about what it is,” said Chris Kraft, the Department of Homeland Security’s deputy CTO for AI. “The information that comes out may be accurate, or it may not be.”

AI governance can also address risks like harmful bias, but determining measures can be challenging, according to Martin Stanley, strategic technology branch chief for CISA.

“There’s not a one size fits all,” Stanley said. “It depends on your stakeholders, the number of people, and the way that they interact with each other within that particular context of use.”

Creating an AI-Ready Workforce

For upskilling employees to better use AI, change can be painful, but investing in training can contribute to a learning culture that encourages experimentation, said CISA CIO Robert Costello.

“We really want to push people as leaders to make change,” Costello said. “You’ve got to have a fast culture, I believe in, and learn from the lessons of what worked. You have to invest in training.”

As the demand for AI expertise in government operations grows, federal leaders are exploring new recruitment strategies to widen the tech talent pipeline in diverse areas, such as looking for adaptive individuals with hard and soft skills.

“We need people who can translate technology to business value because if you can’t sell it to leadership in a way that they understand, they won’t get it,” said Air Force Research Laboratory AI Lead Amanda Bullock. “With a lot of new innovations coming out, we need people to think creatively, which is a shift in traditional government.”

Bryan Lane, a member of the Stimson Center’s Loomis Innovation Council, pointed out that there are also problems with buying AI solutions in government. According to Lane, responsible acquisition through evaluations of assurance can help prevent buyers from being locked into technologies that become outdated.

“People are afraid of new technology and what the risks are, implementation costs, and how fast the industry is changing,” said Lane. “Ultimately, the challenge at hand with acquiring AI is directly related to knowing if what you want to buy today is still going to be useful five or 10 years from now. With the way AI changes on a weekly basis, it’s hard to figure out what exactly to buy.”