States are moving fast on AI. Now, we must measure what matters
sarayut Thaneerat via Getty Images
COMMENTARY | Several are already charting a path forward on how to use the technology, but few have built the systems needed to measure its long-term impact.
As artificial intelligence transforms our economy, more people may turn to government for help with food assistance, healthcare, unemployment and more.
The question is whether the systems they find will be ready for them.
In the past, these systems, many of which were designed for an analog era, have been too slow, too confusing and too fragmented. Now, with a shifting policy landscape and changing eligibility requirements, making these systems work for people is even more critical.
AI can either help government meet this challenge or compound it.
Used responsibly, it can help agencies process information more efficiently, identify missing documents earlier, translate complex rules into clearer guidance and make it easier for people to find and complete the right application. It can also reduce administrative burden for caseworkers and help agencies deliver services with greater speed and accuracy. But if done wrong, it can increase bias, add complexity and reduce trust in government.
Across the country, states are not waiting on the sidelines of this technological shift. They are stepping forward with urgency and a deep commitment to getting it right.
According to new analysis, more states are moving beyond general interest in AI and taking concrete action. They are establishing governance bodies, issuing executive orders, launching secure tests, creating innovation labs and beginning to train public sector workers. Some are testing tools that support document review, eligibility workflows and internal agency operations.
This matters because the next wave of AI adoption will not happen in a vacuum. What takes hold and works in one state will be replicated or adapted in another.
Several states are beginning to shape the trajectory of our government’s use of AI.
Maryland has partnered with Anthropic to deploy an AI-powered agent to help residents navigate the benefits application process online. New Jersey is building on its AI assistant with specialized tools that can expand educational resources, validate documents and analyze high volumes of resident feedback.
Pennsylvania is scaling an AI-driven tool to scan documents for legibility during the benefits application process, helping reduce administrative burden for caseworkers. And Texas has published a robust AI governance framework that provides guidelines and is working to improve data quality across state agencies as it moves from pilots, or smaller tests, to broader projects.
These are just a few examples of what is possible when states approach AI with seriousness, structure and a commitment to learning. They also show that responsible AI adoption will look different across states because public needs, agency capacity, technology infrastructure and policy environments vary widely.
At the same time, few have built the systems needed to consistently measure long-term impact.
That is where the next phase must be focused.
In the year ahead, more states will formalize governance structures to guide their use of AI, expand workforce training and build stronger testing environments. Successful pilots will continue transitioning into operational programs. Procurement strategies will become more sophisticated as states demand stronger evidence of performance, security, accessibility and public value.
The most important shift will be from adoption to accountability. Government leaders should resist the temptation to define success by the number of AI tools launched. The better measure is whether those tools create public value: faster decisions, fewer burdensome steps, more accurate information, better resident experiences and stronger support for the public servants who make government work every day.
Doing so requires clear evaluation frameworks. Government leaders need to know whether AI tools are reducing errors, saving staff time, improving service quality and working effectively across communities. They need feedback loops that include residents, caseworkers, agency leaders, technologists and policymakers. They need systems that allow them to learn, adjust and improve over time.
State governments are already stepping up and embracing responsible AI. The opportunity now is to ensure that AI helps government become more responsive, more effective and more human.
Amanda Renteria is CEO of Code for America, the country’s leading civic tech nonprofit for over 15 years.