Building a foundation for AI-readiness in government

1 day ago 1

Federal agencies are under increasing pressure to do more to meet their mission, with fewer employees and tighter budgets.

Randy Hayes, vice president of public sector at VAST Data Federal, said the Department of Government Efficiency is making it a priority for agencies to adopt artificial intelligence tools more quickly.

“They were forced to efficiency. They were forced to do more with less,” Hayes said. “They’re going to have less budget, they’re going to have less employees. And so, it’s been a forcing function for all of these folks to look at a new way to do things.”

Getting AI-ready

Agencies are also looking at automation tools to maximize the efficiency of their workforce, especially as employees accept voluntary incentives to leave government service.

“Efficiency wasn’t necessarily something that was top of mind for a long time, and now it’s basically law,” Hayes said.

Many agencies see their legacy IT systems as an AI-readiness hurdle. To move from current capabilities to an AI-enabled future, agencies need to focus on interoperability.

“One of the biggest challenges that we’re seeing, getting from A to B, is that there’s this vendor sprawl. So they need to consolidate multiple different vendors, whether it’s a storage system or an AI system or any sort of automation,” Hayes said.

Growing power and cooling requirements for AI infrastructure are also a challenge for agencies.

“If you’re looking at buying a Grace Blackwell cluster from Nvidia, a single rack could be 125 kilowatts of power. Most of the data centers in the D.C. metro area support about nine [kilowatts],” Hayes said. “Trying to retrofit these data centers, from a power and cooling perspective, is going to be a challenge, and it’s going to take a really long time.”

Hayes said that as agencies prepare to get AI-ready, they will need to shift more to infrastructure built by federal systems integrators.

“You’re going to see a lot in the cloud as well, because they have a lot of the scale and infrastructure to handle it,” he said. “We’re seeing from the federal government is they’re trying to get out of the data center operations business.”

Rise of neo clouds and FedRAMP streamlining

Amid all the breakthroughs in this emerging tech space, Hayes highlighted the emergence of “neo clouds” — AI-specialized cloud providers like CoreWeave, Lambda, and G42.

“These are clouds that are just specifically built for AI as a service,” he said. “Their biggest challenge has always been the FedRAMP process.”

That challenge may soon be mitigated. Hayes said new efforts to streamline FedRAMP certification through the General Services Administration’s FedRAMP 20x program could bring approval times for cloud services down to three-to-six weeks.

“Now they’re streamlining the FedRAMP process, there’s been this whole cottage industry to get FedRAMP-certified. These Neo clouds, I think, are going to start coming into the federal government in quite a big way,” Hayes said.

Opportunities to leverage AI across missions

AI is already having an immediate impact on the business of government, especially with its cybersecurity mission.

“We have adversaries that are using a tremendous amount of AI capabilities to attack our critical infrastructure,” Hayes said. “Cyber analytics is a big one.”

AI is also helping agencies provide a better level of customer experience to the public through chatbots and AI agents.

“Whether that’s TSA or CBP — maybe that’s IRS, Social Security, the Department of Veteran Affairs — these are all things that can use agentic AI to help the federal worker provide a better outcome,” Hayes said. “Most of these agents are all going to be prebuilt already. Really, what they need to be able to do is train these agents on their specific data.”

The Defense Department is also using AI for predictive maintenance purposes.

“Something that they’ve been talking about a long time is, ‘I have this Apache helicopter. When is this specific component in this helicopter going to fail? I can be just in time and make sure I have the parts,” Hayes said.

Failing fast to get AI right

Federal agencies have taken a cautious approach to AI adoption, but Hayes said DOGE expects a faster pace for implementation

“The federal government is risk-averse for all the good reasons. But you need to be able to fail fast in this type of environment,” Hayes said. “What they can do is look at lessons learned from the COTS products that are coming from the commercial industries, and who’s actually building these products on the commercial side.”

Hayes urged agency leaders to work with commercial partners who have already deployed these systems at scale.

“You need to be able to fail fast in this type of environment, and the federal government really doesn’t give employees the opportunity to fail, and they don’t give the opportunity to fail fast, especially,” he added.

Hayes said agencies are also taking regulatory steps to vet what kinds of AI tools they’re deploying.

“You are going to have to show your work,” Hayes said. “With all the regulations that are going to come down on AI, you’re going to have to show auditability across who’s touching your data, who’s touching the models, who’s doing what on everything.”

Building for tomorrow, starting today

Even if agencies don’t have compelling use cases for AI right now, Hayes said now is the time for chief information officers to make infrastructure investments that will allow them to become AI-ready.

“You have to buy infrastructure anyway. You have to buy storage and data warehousing anyway. Maybe you’re not ready to go down this agentic AI journey, but if you move your information over and we can contextualize it, then when you’re ready, your data is already ready to roll,” Hayes said.

Read Entire Article