How States Are Turning Artificial Intelligence Into Everyday Government Services
State governments are moving beyond small artificial intelligence experiments and looking at how these tools can support daily work. The biggest challenge is no longer starting a pilot project. It is building a clear plan for what happens when the pilot succeeds.
Start With a Real Problem
Some agencies begin with the idea that they should test artificial intelligence without first identifying the problem they want to solve. That can make it difficult to prove the value of the project or justify expanding it across the government.
A stronger approach begins with a specific need. Leaders should ask what task takes too much time, where errors are common, or which service is difficult for residents to use. The technology should then be chosen to address that need.
A short pilot may be paid for as a one-time project. A full production system requires much more. Agencies must plan for:
- Ongoing funding
- Computer systems and storage
- Technical support
- Cybersecurity
- Data management
- Software licenses
- Performance monitoring
- Staff members who operate and improve the system
Planning for these needs at the beginning helps prevent a useful pilot from being abandoned later.
California Uses AI Across Government Work
California has created a statewide generative AI assistant for employees. Workers can use it to draft reports and other documents, summarize large amounts of information, review complex material, and search policy rules and guidance from trusted state sources.
The state transportation agency has also focused on practical uses tied to transportation operations. Two early projects have moved from testing into regular use.
One system reviews large amounts of road safety information and helps teams find patterns more quickly. Another examines traffic movement and operating conditions so employees can better understand congestion.
The goal is to reduce the time engineers and safety workers spend searching for information. That gives them more time to study problems and develop solutions.
Maryland Sees Strong Employee Adoption
Maryland has expanded access to widely available workplace AI tools across several agencies. About 55% of employees have adopted the tools.
In one case, an agency used the technology to complete work internally instead of hiring an outside vendor. That saved about $400,000.
The state also created a chatbot for a nutrition assistance program. The tool helps residents understand how to apply for benefits and how to receive them. It has handled more than 30,000 conversations, and it resolves questions in about 97% of cases.
Maryland has also built a formal structure for responsible AI use. State law outlines the general governance process, while a senior leadership group oversees the way artificial intelligence is used. The state has issued policies and guidance to help agencies apply those rules.
Colorado Builds Access Into Its AI Program
Colorado first tested an AI productivity tool with more than 150 employees from different agencies. The pilot included workers with disabilities, which helped the state understand how the technology could improve accessibility.
One employee with a vision impairment said the tool had a major effect on her ability to complete daily work. Including employees with different needs helped identify benefits that might otherwise have been missed.
After the pilot, the tool became available across 18 agencies. About 2,000 employees joined at first, and use has continued to grow.
Access is controlled through several steps. Employees request permission, managers approve the request, and users complete training on responsible AI before they begin.
Georgia Focuses on Security and Accessibility
Georgia is providing employees with AI tools, licenses, and training so they can use the technology carefully in their regular work.
The state has also introduced a chatbot designed for a secure government setting. It lets employees search and work with private state data while keeping that information inside the state environment.
Another project uses AI to support accessibility reviews. The system helps agencies examine their services as they work toward compliance with disability access requirements under federal law. The effort supports 18 agencies.
Strong Data and Skilled Workers Are Essential
Artificial intelligence cannot fix poor data. Information must be easy to access, consistent, well managed, and trusted. If records are incomplete or unreliable, the system may produce inaccurate results.
Employee training is just as important. States need to build basic AI knowledge while also involving the workers who understand each agency's daily operations. Many of the strongest ideas come from bringing technology teams together with employees who perform the work.
Agencies also need change management. Simply releasing a new tool does not guarantee that people will use it correctly or regularly. Workers need clear instructions, training, support, and a reason to make the technology part of their routine.
Human Review Must Remain in Place
Artificial intelligence should help government employees, but it should not remove human responsibility. Decisions involving public safety, government services, funding, or individual rights require appropriate human review.
States should also make their systems understandable to the public. That includes explaining where AI is used, what it does, and how people can raise concerns. Agencies should create clear rules for acceptable use so employees know when AI is helpful and when it is not appropriate.
Scaling AI With Care
Higher-risk uses need stronger safeguards. For example, a system that helps decide who receives a public benefit deserves more review than a tool used only to summarize internal documents.
Regular audits can help agencies determine whether a system is working as planned or producing unfair results. These reviews should check for errors, bias, unequal outcomes, and changes in performance over time.
Responsible governance is not limited to the technology office. Cybersecurity, privacy, legal, purchasing, human resources, agency leaders, and program experts all have important roles.
A clear governance process can help good ideas move forward while protecting residents. The most successful state programs are treating AI as a long-term government capability, with careful planning, strong oversight, and clear responsibility from the beginning.