When the new version of ChatGPT was released in November 2022, it felt like someone had cracked open a door we’d been knocking on for years. Suddenly, everyone—from mayors to middle schoolers—was experimenting with a new digital intelligence. 

For the handful of us who had spent years in the civic tech trenches, the question wasn’t just what could this do, but what could this change? Could this be an opportunity to genuinely use technology to transform public institutions and center residents?

That’s how the RethinkAI collaborative was born; a group of self-described civic technologists—policy wonks, professors, and former city officials—who started meeting on Zoom every Thursday to discuss what AI might mean for government and democracy. There was no grant, formal charter, or communications plan, just curiosity and urgency.

The conversations were exhilarating. We discussed everything we’d gotten right in the civic tech movement, open data, digital services, design thinking, and what we missed. Too often, we focused on making government faster, cheaper, and more efficient. But we hadn’t always made it better. We’d built systems that streamlined transactions but didn’t continually deepen trust. We thought AI might finally give us a chance to do that, if we used it right.

That spirit, along with seed funding from the Kresge Foundation and the Chan Zuckerberg Initiative, carried us through two years of interviews, analysis, and our own community-based pilot work, which became our new report, Making AI Work for the Public, published this month. 

Making AI Work for the Public is the most comprehensive look yet at how state and local governments in the United States use AI, pair efficiency with empathy, and move from control to collaboration. 

What Rethink Found Through One of the First Thorough Field Reviews of Local AI

Let’s start with the obvious: AI is already reshaping government, even if most of the action is behind the scenes. Since 2019, over 1,600 AI state bills have been introduced or passed, and in just the first half of 2025, 735 new ones were proposed. Many states are starting to experiment with secure ‘sandbox’ environments. Still, the vast majority of focus is on guardrails—restrictions, audits, and bans—rather than proactive strategies for how AI can make government more responsive to residents. 

Meanwhile, cities are experimenting in more ad hoc ways, including pilots for translation services, public safety dashboards, and community engagement tools. But overall, most governments are still making sense of AI. They’re cautious, curious, and a bit overwhelmed. One official called it “crossing the river by feeling the stones.”

There’s also a new kind of leadership emerging. Instead of mayors or innovation officers leading the charge, this wave is being driven by chief information officers, the same people who used to be quietly keeping the servers running. 

They’re now writing the rules for how AI will be used in public institutions. As one of our interviewees, Erin McKinney of AWS, told us, “CIOs are stepping into leadership roles that are increasingly innovative and strategic, not just operational.” That’s a sea change and bodes well for an enterprise, all-of-government approach to reform.

Learning from Our Own History

For those of us who grew up in the civic tech world, this feels both familiar and new. The civic tech movement gave us open data, hackathons, and a generation of public interest technologists. But it also fell into what we call the efficiency trap. 

We celebrated saving time and money, not realizing that residents weren’t impressed by faster bureaucracies; they wanted different bureaucracies, more human ones, more accountable ones. 

AI, for all its complexity, offers a reset button. But only if we learn from that past. 

The report argues that we need a new kind of governance model, one built on what we call ALT: Adapt, Listen, and Trust

• Adapt means using AI to anticipate future needs, not just react to current crises. In San Jose, the city is using predictive analytics to prevent evictions before they happen. But more than that, it’s using AI to plan for the services—legal aid, rental assistance, housing support—that will be needed to keep people stably housed. It comes down to building public systems that are ready for what’s next.

• Listen means turning AI into a tool for understanding, not just automation. In Boston’s Dorchester neighborhood, we helped build On the Porch, a community-trained language model that lets residents analyze their own neighborhood data. It flips the usual script—residents get to ask the questions and set the agenda.

• Trust means moving from one-way transparency to two-way accountability. In New York, we partnered with CitizensNYC to analyze thousands of community grant proposals using AI, helping surface patterns that can inform future funding and policy. It’s a reminder that data should flow between governments and the people they serve.

The Real Transformation Ahead

If civic tech is about opening up government, civic AI must be about reimagining it. We need institutions that can adapt quickly as technology changes, listen to communities in real time, and earn trust through accountability and results.

That’s not just a technical challenge, it’s a moral one. The public sector is where society negotiates fairness, equity, and belonging. If we let AI be only about cost savings or automation, we’ll miss its real potential to rebuild legitimacy in our institutions.

The report describes this as a move from governments to governance; a broader ecosystem that includes philanthropy, higher education, nonprofits, and community groups all playing a role. Philanthropy can fund what governments can’t risk. Universities can train and evaluate. Communities can ground the work in lived experience. And coalitions like ours—RethinkAI—can keep the conversation honest, ambitious, and human.

We started as civic technologists meeting on Zoom, just trying to make sense of the AI wave. Two years later, we’ve come to believe that if we aim AI at efficiency, we’ll get faster bureaucracy. But if we seek Adaptation, Listening, and Trust, we’ll get something far better: a public sector that people believe in.

Read the full report: Making AI Work for the Public
Authors: Neil Kleiman, Eric Gordon, and Mai-Ling Garcia for RethinkAI and New America’s Technology and Democracy Program