Where US agencies are finding measurable value from AI

The report reveals a widening gap between enthusiasm and execution. While 57 per cent of agencies are actively exploring and learning about AI, only 16 per cent are piloting small projects, and just 1.6 per cent report broad deployment across departments.

According to respondents, adoption is slowed not by scepticism, but by structural constraints, including security and privacy concerns, unclear governance and policy guidance, legacy systems, and limited staff capacity. 

Main findings include:

• Interest outpaces deployment: 57 per cent of agencies are actively exploring and learning about AI, but only 16 per cent are piloting small projects, and just 1.6 per cent report broad deployment across departments
• Operational AI delivers earlier value than generative tools: respondents see the most promise in automating high-volume, rules-based workflows rather than deploying standalone generative copilots
• Time-savings is the top success metric: more than two-thirds of respondents identified productivity and hours saved as their primary desired outcomes, ahead of cost savings or technical sophistication
• Financial workflows lead early adoption: top near-term use cases include procurement RFP generation (29 per cent), forecasting and scenario planning (27 per cent), and grants research and matching (25 per cent). 

The report also emphasises that agencies want to adopt AI, but many lack the foundational conditions needed to scale safely. Adoption is slowed not by scepticism but by infrastructure, governance, and capacity limitations. 

The report finds primary barriers to AI adoption include: 

• Protecting sensitive resident and financial data 
• Lack of clear federal or state AI governance guidance 
• Legacy technology constraints 
• Limited staff capacity and organisational bandwidth.
  

Despite these challenges, the report highlights a clear path forward. Agencies that succeed with AI tend to start small, focus on workflows rather than tools, and build governance in parallel with experimentation. In many cases, the fastest progress comes from evaluating AI capabilities already embedded within trusted govtech platforms, reducing risk while delivering immediate efficiency gains.

“The greatest impact of AI is coming from practical applications that strengthen the backbone of government operations – procurement, budgeting, grants, and payments”

It concludes that AI is rapidly becoming a foundational layer of public sector infrastructure, much like cloud computing and cybersecurity before it. Agencies that define success early – using practical metrics such as time saved, accuracy improved, and workload reduced – will be better positioned to scale AI responsibly over the next 24 to 36 months. 

“Public sector teams don’t need to deploy AI everywhere overnight,” said Amburgey. “Real progress happens when agencies rely on trusted platforms where AI expertise is already thoughtfully embedded into essential workflows. With AI built directly into the systems governments depend on, agencies can reduce risk, gain efficiency, and better serve their communities.”

The survey of 58 public-sector finance and operations leaders conducted at Euna Solutions’ 2025 Eunaverse user conference (September 2025). 

Read the full State of AI in Public Sector report here.