Staff shortages block AI adoption across US government

GovLoop has published research showing that staffing shortages are the main barrier to artificial intelligence adoption across the US government. The findings are based on a survey conducted with Granicus and Carahsoft.

The survey of 121 workers across federal, state and local government found that nearly 60% cited a lack of staff capacity as a leading obstacle to progress. By contrast, only 17% said a lack of viable use cases was a serious problem, suggesting agencies are constrained less by ideas than by the workforce needed to put them into practice.

Security emerged as the most common concern. Some 65% of respondents said the security of the AI model and the data behind it was a top worry, while just over half said poor interoperability between systems was limiting their ability to make AI work as intended.

The figures suggest the public sector debate has shifted from whether AI may be useful to whether agencies have the resources and controls to deploy it. They also show that practical issues inside organisations weigh more heavily than uncertainty over the technology itself.

State and local

A subset of 63 respondents from state, city and county government broadly reflected the national picture, but with notable differences. State and local workers reported greater confidence in their organisation's ability to manage AI risk, with nearly 70% saying they were at least moderately confident, compared with just over 60% across the full sample.

They were also less likely than the national average to cite unclear leadership direction as a barrier, at 32% versus 39%. Similarly, fewer said a lack of proven results was holding back adoption, at 30% compared with 37% nationally.

Yet the same group reported greater pressure in other areas. Procurement hurdles were cited by 43% of state and local respondents as a barrier to modernising services, compared with 35% nationally. Nearly 73% also described moderate to high pressure to deliver faster and simpler services for residents, against roughly 62% across the broader sample.

That gap suggests local agencies may feel both closer to public demand and more exposed to administrative delays when trying to respond. The data indicates procurement processes are a more acute obstacle for these bodies than for federal agencies.

Governance gaps

The research also highlights immature governance arrangements in many state and local organisations. Only 8% of respondents from state, city and county government described their agency's AI governance as centralised and mature.

Nearly 43% said governance rules were still being written, while 38% described their approach as informal or ad hoc. That leaves a large share of agencies operating with incomplete structures at a time when AI use is attracting greater scrutiny over accountability, risk and data quality.

The findings suggest confidence in managing AI risk does not always translate into formal governance systems. In practice, some agencies appear to believe they can handle the risks while still lacking settled internal rules on responsibility and oversight.

Operational risks

Respondents from state, city and county government were also asked what could happen if their agency failed to modernise resident service access over the next two years. More than 52% said they expected rising operational costs.

Just over half expected larger backlogs and slower response times for residents. The same proportion anticipated higher staff burnout and turnover, while nearly 48% said failure to modernise could make it harder to meet mandates or compliance requirements.

Only 36.5% said there would be no significant consequences. That suggests many public sector workers see inaction not as a neutral choice, but as a route to higher costs and worsening service pressures.

Karthik Anbalagan, General Manager of emerging technologies at Granicus, offered his assessment of the state and local findings.

"People at the state and local level aren't wondering if AI can help them. They already know it can. What's interesting is they're more confident managing the risk than their federal counterparts. What's holding them back isn't confidence. It's procurement friction and governance rules that haven't caught up yet. And the survey is clear that standing still isn't free. Agencies that wait are telling us they expect additional costs, longer waits for residents, and burned out staff," Anbalagan said.

Anbalagan also pointed to trust and data provenance as central concerns in government use of AI.

"If you don't know where the answer came from, you don't trust it. Agencies need to know their AI is only using information they've approved, not pulling from wherever it finds something online. If the information behind it is out of date or messy to begin with, AI just repeats those same mistakes faster," he said.

He added that AI deployment depends on more than technology alone.

"You can't fix a people problem with just technology alone. The agencies getting ahead are training their staff, cleaning up their data and being clear about who owns what. That's where it starts, and right now state and local government has the confidence to do it. What it needs is fewer procurement roadblocks and clearer governance to match," Anbalagan said.