New Jersey, Pennsylvania and Utah have emerged as national leaders in building the governance, workforce training, and technical infrastructure needed for artificial intelligence adoption in state government, according to a new assessment released Tuesday by Code for America.
How States Are Implementing AI
The leading states have all prioritized learning, leadership, and technical infrastructure in their AI readiness strategies. Utah established the nation’s first dedicated AI policy office last year and created an “AI Learning Lab” to exchange knowledge among developers, communities, researchers, and government. New Jersey established a Chief AI Strategist role to support the state’s use of generative AI to deliver benefits and services. The state also provides its employees access to GenAI tools in a secure environment through its NJ AI Assistant. Pennsylvania has collaborated with Carnegie Mellon University to support AI research and implementation in the state, and has created a pilot program that provides employees with enterprise ChatGPT licenses to help with common work tasks. All three states have also provided state employees with training in AI skills through partnerships with InnovateUS.
Many states are also using AI to develop practical applications that improve service delivery, make government operations more efficient, and deepen engagement with residents. Earlier this year, California announced an experiment with AI-powered deliberation tools to gather ideas from residents about how to support communities impacted by the 2024 Los Angeles wildfires. In 2024, New Jersey’s AI Task Force developed a custom AI-powered research toolkit, combined with an online engagement with thousands of workers, to develop recommendations for how New Jersey should prepare for AI’s impact on the economy and workforce. Texas’s Office of the Attorney General has created an AI-powered search tool which helps child support field case workers query case information and legal documents more efficiently, speeding up their time-intensive research process. In a similar vein, the Massachusetts Department of Transportation has developed a chatbot for design engineers to speed up infrastructure planning by streamlining manual searches. The AI tool was developed in collaboration with students through Northeastern University’s AI for Impact co-op program.
How Code for America Analyzed AI Readiness
The Government AI Landscape Assessment evaluated all 50 states across three dimensions: leadership and governance, AI capacity building, and technical infrastructure and capabilities. The Beehive State, the Garden State, and the Keystone State demonstrated advanced capabilities across multiple dimensions, setting them apart from the remaining 47 states, according to the analysis. Half of all states were described as “Established,” – meaning that their AI implementation is “mature” with “systematic approaches and demonstrated effectiveness” – with the remaining states either at the “Early” or “Developing” stages of their AI readiness journey.
"Across 50 states, I think people are realizing that AI is something that needs to be considered in service delivery," Jenn Thom, Code for America’s lead data scientist, told StateScoop. "And while there are some states who have made progress more quickly, I think those states can serve as a model for states who are just dipping their toes in or starting to explore."
The Capacity Gap
The assessment found that capacity building remains the least developed area nationwide, with the majority of states still in early stages of AI training and workforce development. Most states have taken some steps to organize AI-related leadership through task forces or assigning responsibilities to existing officials, but few have appointed dedicated Chief AI Officers.
The need to build institutional capacity for the public sector to truly capitalize on the opportunities presented by GenAI was also highlighted in a recent article by Tiago Peixoto, a digital government expert at the World Bank. “GenAI deployment cannot be reduced to technical upgrades,” Peixoto writes. “It requires confronting institutional arrangements that benefit from complexity. Without this, in certain contexts, even advanced tools risk entrenching existing inefficiencies while adding new layers of technological dependency.”
Echoing Peixoto’s warning, the Code for America report shows that the states leading in AI readiness are those that have moved beyond using AI to patch holes, and are instead building comprehensive governance frameworks, investing in workforce development, and establishing dedicated leadership structures to support public agencies in using AI. It shows us how far states have come in just a few years – and much work is left to be done.
