The people problem behind the government’s AI ambitions

Late last month, President Joe Biden issued an executive order on artificial intelligence with over 100 specific tasks across more than 50 agencies, directing them to harness their regulatory powers in the realm of AI, create safety testing standards for systems and more.

The administration now faces the challenge of beefing up the government’s AI workforce by training existing feds and hiring experts. 

Both efforts are table stakes for the government to execute new policies meant to shape the future of the technology and capitalize on AI to improve government operations, experts and stakeholders say. And implementation is starting soon. Many executive order due dates fall within a year, with the 2024 election and potential administration change looming.

“One of the biggest barriers, certainly the one that we hear most about, is in the workforce challenges,” said Conrad Stosz, director of AI in the Office of Management and Budget’s Office of the Chief Information Officer, at a recent IBM event. 

The executive order mandates a National AI Talent Surge into the federal workforce, requiring the government to identify priority mission areas in need of AI talent, create a task force for hiring and more. OPM’s director told Nextgov/FCW previously that the agency plans to draw on its experience in a hiring surge for the Bipartisan Infrastructure Law.

The recent executive order directs OPM to pull various federal HR levers to enable flexibilities around the government’s civil service requirements, like excepted service appointments, pay flexibilities and pooled hiring across agencies.

Agencies will leverage existing talent programs like the White House’s U.S. Digital Service to pull in AI-related talent, and eventually, OPM is expected to issue guidance on skills-based hiring for AI, tech and data talent, a move experts say could open the government up to candidates from nontraditional backgrounds.

The executive order also calls for new AI training programs. OMB and the General Services Administration have already hosted a series of several voluntary training sessions for feds with Stanford University, Stosz said — with over 8,000 feds signing up to participate across over 85 agencies.

Among the requirements in the executive order, workforce policy generated the most to-do items, according to a new Stanford University Institute of Human-Centered AI tracker of such requirements. 

The first-place ranking reflects not only the scale of the need in government, but also the complicated federal HR environment — what Stanford researchers called a “vast morass of confusing hiring authorities and procedures.”

The government has already been hard at work of late hiring techies in general. The public sector’s push to scoop up laid off and dissatisfied technologists from industry has seen huge interest — a recent hiring fair with state, local and federal agencies had over 3,000 registrants and a 1,500-person waiting list, according to organizers — but the actual hiring process can be confusing for job seekers and hiring managers alike.

“There's technologists who really want to make an impact, and they don't know how to navigate hiring,” said Jennifer Anastasoff, USDS alum and founder and executive director of the Tech Talent Project, a nonprofit that’s part of a coalition on government tech hiring. “We need to do better.”

The backdrop for the current AI talent surge, Stanford researchers say, is that OPM is behind on existing requirements from the AI in Government Act, passed nearly three years ago as part of 2021 appropriations law. It required OPM to create an occupational series for AI — or update an existing one — and estimate the number of current AI feds while forecasting the need for the future.

OPM has released core skills and competencies related to AI already, also required by the law, but the researchers ask, “With OPM over two years behind on many requirements, how real will this hiring surge be?”

An OPM spokesperson told Nextgov/FCW that it has issued a data call to agencies for estimates on the number of federal AI positions and forecasts for coming years, and the personnel agency has gotten numbers back. OPM has also done an occupational study with the White House’s Office of Science and Technology Policy, the spokesperson said, although it hasn’t publicly released findings.

Overall, less than 40% of the requirements contained in two Trump-era executive orders on AI and in the 2020 AI in Government Act have been completed, according to publicly available information, a paper published by Stanford researchers in August states, although the status of many requirements couldn’t be verified by the researchers.

Researchers cited “a resource shortage, a leadership vacuum and a capacity gap,” supercharged by policy ambiguity, as feeding into the lag.

Part of the delay was due to competing priorities, including pandemic response, said Suzette Kent, former federal CIO during the Trump administration and current advisor to the stackArmor Risk Management Center of Excellence. How the release of the policies aligned, or didn’t, with budget cycles was also a factor, although Kent said “progress was made” in areas like agencies’ AI inventories and data. 

Now, “a key question will be the budgetary commitments to make this work,” Daniel Ho, one of those researchers and a law professor at Stanford University, told Nextgov/FCW via email, about the implementation of the latest EO writ large.

“One of the first things I’d do is [ask], ‘Do I have the right people?’” confirmed Maria Roat, former deputy federal chief information officer who is also a stackArmor advisor. “‘How do I hire people in a limited budget situation under a [continuing resolution]?’ It’s hard.”

Congress passed a two-tiered continuing resolution earlier this month to fund agencies through January and February, but what will happen when lawmakers reach those deadlines is unclear.

An OMB spokesperson told Nextgov/FCW in a statement that “the administration will continue to work with Congress to provide additional funding.”

As for the stakes, “federal agencies have a distinct responsibility to take this issue seriously… given the importance that their missions have in the lives of the public,” said Stosz. “We just can’t afford to get this wrong.”