Trump administration hopes AI can mitigate staffing losses, federal CIO says

The Trump administration is “100%” looking to artificial intelligence to make up for the loss of thousands of employees across federal agencies, a top Trump administration tech official told Nextgov/FCW in a recent interview.

“I think [AI] is the number one thing that is going to help people mitigate the staffing shortages,” said Gregory Barbaccia, the chief information officer for the U.S. government, a position with oversight of the federal government's technology that’s housed within the Office of Management and Budget. Barbaccia, who's also been identifying himself as the federal chief AI officer, was a former longtime employee at data analytics company Palantir.

In one of his first interviews after about eight months on the job, Barbaccia spoke with Nextgov/FCW about the administration’s vision for AI and how it intersects with the White House’s push to shrink the size of government through both layoffs and voluntary incentives. Since Trump took office in January, over 148,000 civil servants have left the federal workforce. 

The biggest challenge in his role, Barbaccia said, “is how to do more with less — how to still have the level of efficiency, following up on your agency’s mission for the taxpayer, operating in an environment of lean staffing.”

“AI is a huge part of that, so that’s a major focus for me,” he added.

Another priority is data, Barbaccia said.

Enterprise data platforms — central systems connecting data from across an organization — are “what’s going to power AI and advanced analytics to actually get people [sic] more efficient,” he said. “AI is going to be a huge part of this, all these manual, repeatable tasks.”

“What AI needs to be doing,” Barbaccia continued, “is collating and collecting this data that a human needs to make a decision and getting it to them in a way that makes sense, in a way that's quick, so they're able to make a decision with all the information they need as quickly as possible.”

Barbaccia’s remarks come days after the launch of USAi, an evaluation platform meant to help agencies adopt AI models from Anthropic, Google, Meta, Microsoft, Amazon and OpenAI, some of which recently cut deals with the government to offer their software for nominal fees. 

Proponents of AI say it could aid the government in making sense of its vast troves of data, support federal employees and offer real-time help to people seeking government assistance without them having to call agency phone lines.

“How do we use tech to take the excellent humans that we still have and make them as efficient as possible?” Barbaccia said of AI. “How do we get them doing what they were doing before, but doing it way faster and better?”

The “holy grail” is “if we're automating more things that can be done by a computer, what are humans doing that they could not do before?” he added.

Max Stier, the president and CEO of the Partnership for Public Service — a nonprofit focused on good government — told Nextgov/FCW in a statement that “federal leaders, as stewards of public funds, have an obligation to leverage cutting-edge technology to address our collective problems.”

“The federal government can and should take concrete steps to enhance the technical capacity of its workforce to fully take advantage of all the possibilities AI has to offer,” Stier added. “The civil service needs to evolve, but it cannot be replaced by AI.”

Those more critical of AI have cautioned about its potential to cause harm or automate bias at scale, a point especially relevant as the Trump administration has targeted so-called “woke” AI. Critics say that cracking down on AI tools deemed to be too woke could chill both free speech and AI-focused risk management in government agencies.

There’s also questions about the reliability of the technology — generative AI models can produce plausible-seeming but ultimately false outputs called hallucinations — and the fact that many agencies have lost staff that would presumably be called upon to implement new AI solutions.

“It's very early days for the use of AI in government services. But what we do know is that the approaches that work well are those where the employees drive the process of integration in a bottom up and grounded way,” said Suresh Venkatasubramanian, the director of the Center for Technological Responsibility, Reimagination and Redesign at Brown University who did a stint in the Biden administration. 

“What doesn't work is firing government workers who have expertise and then trying to use AI to replace their years of knowledge and experience,” Venkatasubramanian said. “And what certainly doesn't work is forcing the remaining workers to be constrained to use only AI tools that conform to Administration speech requirements. Fast is not synonymous with better, and in fact is often the opposite of it.”

As for the data to power AI, the Trump administration is already pushing agencies to more easily share data, and its Department of Government Efficiency has been at the center of many headlines and lawsuits over its hoovering up of government data. 

In the spring, Wired and CNN both reported that DOGE is building a “master database” with data from different agencies to track undocumented immigrants. The New York Times also reported on the expanded use of Palantir across the federal government.

House Democrats also issued a whistleblower-informed report in April alleging that DOGE was building a cross-agency database of sensitive information from the IRS, SSA and other agencies.

Barbaccia maintained that he doesn’t know of such an effort. 

“I can’t find it for the life of me,” he said. “I have no insight into any sort of master database going on. I asked agencies at IRS and SSA. They don't have anything they're building outside of their internal data. I've asked the vendors who are in the news, too. They're not tracking what the news is about.”

“I think it’s people misunderstanding what different vendors do,” he said.