Can AI reshape how states manage vendors? Here’s $10M to find out

The Center for Civic Futures, a nonprofit that in December announced millions of dollars in grant funding aimed at propagating the most effective ways to improve state governments’ benefits systems, on Monday announced a second, $10 million round of funding, through its Public Benefit Innovation Fund.

This round is joined by the nonprofit Recoding America Fund, an initiative co-launched last October by Jennifer Pahlka, who founded and formerly led Code for America, the civic tech nonprofit that does the similar work of helping governments develop slicker and more effective digital tools. For the next five weeks, the groups will collect applications from nonprofits, early-stage companies, academic institutions and government agencies that want money to fund technological fixes to the key challenges facing benefits programs like Medicaid and SNAP: “reducing administrative burden, closing enrollment gaps and improving the experience of both benefit recipients and government staff,” according to their press materials.

Replacing government’s most irritating bureaucratic processes with user experiences more closely resembling an Amazon purchase or a Domino’s order has been on the agenda of a certain variety of civic-minded technologists since at least the first Obama administration. But one of the many unintended effects of the second Trump administration has been to bestow that project with new urgency. New reporting requirements for benefits programs included in last year’s “big and beautiful” budget reconciliation bill are propelling states to make IT upgrades.

But these civic tech organizers worry that states will follow a predictable pattern of funneling too much responsibility and too much money to large government technology companies with familiar names. An editorial co-authored by Pahlka and Cassandra Madison, the Center for Civic Futures’s executive director, notes hopefully that artificial intelligence tools may put states “in the driver’s seat, making them better informed buyers, more capable builders and better stewards of outcomes.” But they warn, too, of an alternate future, noting that government agencies’ procurement processes aren’t designed to build the public sector’s 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.”

The two groups’ organizers said this round of funding is unique because each will focus on one half of the challenge, as they see it. The Center for Civic Futures will focus more heavily on AI tools that can assist case workers and benefits applicants. The Recoding America Fund will target the institutional issues that might confound the adoption of fancy new AI tools. The good news, according to Madison, is that governments are interested — she pointed to the 400-plus applications her organization received in last year’s $8.5 million funding round.

Recent months of “deep listening” sessions led by the center, with case workers, chief AI officers and other government IT leaders, revealed three issues that should be addressed. The first, she said, is “back-end transformation”: “Pretty much every program faces backlogs. And very difficult, very manual processes.” Madison said caseworkers want to help people, yet they still spend “an awful lot of their time on paperwork.” She hoped that testing emerging technologies, like the latest class of generative AI tools, will reduce those workloads — after projects funded by these grants reveal what works and which ideas aren’t worth pursuing further.

The second issue, she said, is data. If Medicaid and labor agencies can more easily share data, it could mean fewer eligible recipients will lose coverage because of bureaucratic snags.

The third issue is “enabling infrastructure,” the specialty of the Recoding America Fund. “There’s parts of government that aren’t digitized yet,” Madison said, and “there’s legacy processes and systems that get in the way of states, territories and tribes being able to experiment with these new technologies.” The new funding is to “figure out what are those walls that people are going to hit and how can we experiment with ways to unblock, and allow for smart innovation.”

Unlike last year’s funding, this round will distribute money for two categories. Early-concept ideas, to be tested as prototypes for as long as one year, can receive awards of less than $500,000. Concepts ready for real-world testing, for up to two years, can receive funding of up to $2 million. Funded projects get “technical assistance, expert coaching, access to government leaders and end users, and visibility with funders and policymakers,” according to a press release. The awards are to be announced in September.

Robert Gordon, the Recoding America Fund’s executive vice president for state initiatives, said this work is necessary because governments continue to operate from an outdated mental model of how technology works: “They think of building technology as analogous to building a bridge,” a process in which thousands of requirements are bid out in a lengthy procurement, and “then they’re going to expect the vendor to do all the work. Under that model, you end up with a vendor first of all that isn’t super responsive to user needs,” but that “knows and manages everything, with often a very thin layer of state oversight.” 

This is a decent model in a world where nothing goes wrong. In practice, Gordon said, the state depends on its vendors for everything, from fixing buggy code and benchmarking system efficacy to learning how much it should be paying for such services. “That is the strategy that has left states disempowered relative to vendors,” he said. Though he admitted that “vendors are really important. None of us imagines a world in which they disappear. The question is: Do states have the capability to manage them differently? And we think AI has the potential to help states manage them differently, at much lower cost and much more effectively than before.”