Guide
How agencies use AI, what official policy says, where the risks sit and how to start. An AI guide for government teams and the contractors that serve them.
Updated 7 min read
The short answer
AI in government means public agencies using software that reads, writes, predicts or sorts to do their work faster. It handles tasks like processing forms, spotting fraud and answering requests, while people keep the decisions that matter.
Key points
AI in government covers any use of machine learning or generative AI inside a public agency. Artificial intelligence in government sector work spans federal, state and local offices.
Most of it is plain office work: reading documents, sorting requests, drafting text and finding errors.
GAO looked at generative AI use cases at selected agencies. It found 61 percent fall under mission-enabling functions, such as finance, human resources and procurement.
1,110
AI use cases at 11 agencies in 2024, up from 571
282
generative AI use cases in 2024, up from 32
~60%
of use cases still in pilot or pre-deployment
Federal government AI use is growing fast. At the 11 agencies it reviewed, the Government Accountability Office found that reported AI use cases nearly doubled, from 571 in 2023 to 1,110 in 2024.
Generative AI grew even faster, from 32 to 282 use cases. Brookings notes that nearly 60% of all use cases are still in the pilot or pre-deployment stage.
Large agencies contributed 76% of reported use cases in 2025, up from 69% in 2024.
$4B+
fraud and improper payments prevented or recovered in a year
$1B
recovered with machine learning on check fraud
AI applications in government fall into a few groups. Machine learning in government spots patterns, and generative AI handles text.
The Treasury reported that machine learning helped identify check fraud, resulting in $1 billion in recovery.
Treasury also credited risk-based screening with $500 million in prevention. That was part of more than $4 billion in fraud and improper payments the agency itself reported preventing or recovering that fiscal year.
AI in the public sector also runs at state and local level, under its own rules. Governors and legislatures have issued orders on generative AI and set up task forces, as tracked by Government Technology.
The Municipal Research and Services Center notes that AI brings legal risks for local offices, such as confidentiality and data privacy concerns, inaccuracies and bias.
AI in government services that reach residents directly, like chat tools, needs the most care.
94
government-wide AI requirements identified
$1.9B
requested for AI research and development, fiscal 2024
Executive Order 14179, signed January 23, 2025, states that the policy of the United States is to sustain and enhance America's global AI dominance. It also revokes certain earlier AI policies.
The Government Accountability Office identified 94 AI-related requirements that were government-wide or had government-wide implications. Agencies also requested $1.9 billion for AI research and development in fiscal year 2024, GAO reported.
Procurement rules shape what vendors can sell. OMB memo M-25-22 requires agencies to update acquisition policies and share best practices, according to Digital Government Hub.
It covers the full lifecycle, from risk identification and market research to contract closeout.
A law firm's summary adds that agencies must bar the use of non-public agency data to train public or commercial AI without agency consent.
10 of 12
agencies said current federal policy could slow adoption
Agencies name their own obstacles. Officials at 10 of 12 agencies GAO selected said existing federal policy, such as data privacy policy, could present obstacles to adoption. Government errors reach real residents, so safeguards matter.
The NIST AI Risk Management Framework gives a structure: govern, map, measure and manage, as Brookings describes it. Agentic AI in government, where software completes several steps on its own, calls for even tighter review.
New America's field review offers a useful rule: lead with problems, not tech. It also notes that basic productivity gains are far easier to achieve than ambitious transformations.
Pick one repetitive process, measure the hours it takes, and test on real work. Our government automation work follows that path for agencies and the contractors that sell to them.
Our data
We asked ChatGPT, Gemini and Perplexity 10 buyer questions about AI for government contractors and agencies. They named 90 companies, and 69% appeared in only one of the three.
| Criterion | ChatGPT | Gemini | Perplexity |
|---|---|---|---|
| Companies named | 43 | 38 | 47 |
| Sites cited most | govdash.com and deloitte.com (3 of 10) | civio.ai (5 of 10) | govdash.com and technomile.com (5 of 10) |
Nexum measurement, October 20, 2026, via DataForSEO: 10 buyer questions to each AI, US English, 30 answers in total. 10 companies appeared in all three. Google showed an AI Overview for 7 of the 10 questions.
FAQ
Final step
You leave knowing which processes to automate first and what each one costs you today, whether we work together or not.
You can keep paying for manual work, or find out this week exactly what it costs you.
What happens on the call
Before
You answer 3 quick questions when you book, so we come prepared.
On the call
We find the manual work that costs you most and put a number on it, live.
Decide
An honest read on whether we're the right fit. If we're not, you'll hear it on the call.
Once you book, we look at your site and tools before the call, so the 30 minutes go to your numbers, not introductions.
Prefer to write first? Email us directly: contact@nexumautomations.com
