Every state in America is rushing to spend millions of dollars on Medicaid updates that will not work properly.
In an attempt to meet new work reporting requirements mandated by the One Big Beautiful Bill Act, states will direct funds to the same small set of companies that have defined government technology for decades. Many will fall victim to the same predictable mistakes. The losers will be the taxpayers who foot the bill for these changes, the frontline workers who have to navigate yet another round of technology changes that don’t make their jobs easier, and the Americans who can’t meet the new requirements.
We know this from our experience over the past two decades — serving in state and federal government, supporting government at Code for America and Tech Talent Project, and now in our current capacities at the Recoding America Fund and Center for Civic Futures.
The last decade of civic technology — work involving thousands of extraordinary people — has improved the government’s ability to deliver benefits. In pockets, this work has improved services dramatically. But it has not fundamentally changed the way states engage legacy vendors, what they pay or the results they get. The U.S. just sent four astronauts farther from Earth than any humans have traveled before, but we still can’t get benefits delivery systems to work.
AI could change that.
Artificial intelligence tools are making it dramatically cheaper and faster to write software and make sense of data. States can now build, measure, learn and iterate like never before. Under the right circumstances, small state teams with AI tools could quickly learn what it takes to implement a new policy, draft code updates, design and analyze tests, and make changes to live systems in days, rather than months. Even for purchased software, they could build critical tools on top: better workflows for participants, dashboards for decision-makers and bridges between disparate systems. In short, AI can put state governments in the driver’s seat, making them better informed buyers, more capable builders and better stewards of outcomes.
This is already happening, though not in American governments. In the private sector, companies are discovering that seven-figure service contracts can be replaced with in-house builds that work better, for a tiny fraction of the cost. In Singapore, government staff with these tools are seeing productivity gains as great as 22x.
The question is not if AI can unlock progress, but whether governments will adopt emerging technologies in ways that make them more capable or more dependent. The way government buys and manages technology today does little to build capacity. The barriers aren’t just technical — they are structural and operational. States may not have access to the code they paid for. Internal review processes are not yet designed for the rapid, iterative improvement that AI tools make possible. Procurement rules are too slow and rigid to support experimentation. And many agencies lack the workforce, culture or leadership to try something new. The people who would need to authorize change — general counsels, procurement officials, agency directors — are embedded in cultures that treat change as a prohibitive risk, or leave their positions before they can make meaningful progress.
Tackling these binding constraints gets little investment. That’s why the Recoding America Fund and Center for Civic Futures are teaming up on an open call for proposals through the Public Benefit Innovation Fund. Run by Center for Civic Futures, PBIF has a strong track record of funding emerging tech-enabled tools that improve public benefits delivery: applications that help caseworkers, connect people to services and modernize backend systems. The Recoding America Fund invests in setting the conditions that determine whether those tools can actually be adopted to drive impact.
Through this open call awarding up to $10 million, we’re seeking proposals that address both sides of the problem. We want teams that are using AI in promising ways to deliver government services better and improve lives — to show what a new operating model can look like in practice. And we want teams tackling the structural and operational barriers that prevent governments from building with AI in the first place.
A brilliant prototype that can’t get through procurement won’t help anyone. An AI-powered case management tool that a state can’t maintain without a vendor hasn’t changed anything fundamental.
The One Big Beautiful Bill Act demands change on a timeline incompatible with business-as-usual procurement. State budgets are already under strain. And workforce displacement may soon stress unemployment insurance systems, risking the kind of backlogs we saw during the COVID-19 pandemic. Who knows what other kinds of disruptions are around the corner? Government’s current operating model cannot meet these demands at the speed that’s required.
The tools to change that model are arriving alongside the pressure to use them, but we have a narrow window to act. If we only use this moment to sell government a new product, we’ll have wasted the moment. If we use it to help government build capacity to guide needed work independently, we will achieve lasting change.
Jen Pahlka is the co-founder and board chair of the Recoding America Fund.
Cass Madison is the executive director of the Center for Civic Futures.