AI has the potential to transform government operations, but there are barriers slowing public-sector adoption, including limited data skills and concerns around data privacy. Sandbox environments, however, offer governments a method through which to responsibly explore potential solutions powered by AI.
Social Finance launched its hub last year, building on earlier work with an AI Learning Lab at the Benefits Data Trust which has since closed down. Its intent is to help governments navigate the rapidly changing AI landscape and its impacts on government procurement, Abby Silverman, the organization’s associate director of communications and public relations, said.
Her public-sector experience, Silverman said, gave her insight into a common government challenge: “Government practitioners are often there to make the world a better place and to do their jobs because they came there for some mission … but they don’t understand the contracts.”
Some government officials have a limited understanding of technology and data, which can result in contracts that fail to maximize the benefits for public sector and the people it serves; AI, Silverman said, is adding a new layer to this issue. The AI sandbox created through the hub allows governments to experiment with AI before the procurement cycle — and even build AI solutions. The ultimate goal, she said, is to empower governments with knowledge prior to signing a contract with a vendor. The initiative highlights “outcomes-based contracts,” which allow governments to keep vendors accountable for claims they make about what their technology can do.
“We can’t oversell in the government space, because these tools are being used to affect folks’ lives,” Silverman said.
The first cohort was a pilot, she said, and participants were largely selected based on existing relationships, and via requests to state and local governments. Each cohort is meant to last approximately six months; the goal is for the second to be active by spring.
One government that participated in the first cohort is the city of Alexandria. Its CIO Vanetta Pledger said the city got involved with the hub to explore potential AI use cases. Officials decided to focus on the Department of Code Administration, as the related data was not legally protected or sensitive. Staffers there do building code reviews to ensure compliance with building requirements, but the process can be time-consuming.
The goal, Pledger said, was to see how AI could help staff more easily access information to process these reviews more efficiently — not to replace staff, she said, arguing that one role of leadership is to help employees understand how AI technology can “complement” the work they’re doing.
Quality data powers successful AI implementations, but government data maturity and readiness levels vary. Pledger said one of the challenges she hoped the hub would help her solve was assessing and improving the city’s data quality.
Local governments have many different data sets, Pledger said, so using AI to help understand existing trends, needs and possible remedies with city-designed solutions can support operations by automating routine tasks. This can allow staff to focus on more complex tasks and community interactions. The hub enabled data aggregation, letting the city securely synthesize multiple data sets, she said, and answer complex questions about that data.
“And then I think it empowers you to sort of challenge and use your critical thinking to better understand how to address certain situations,” she said of the hub.
Hub participation entailed governments identifying several challenges they thought AI could help solve, and then collaboratively thinking them through with the Social Finance team. Participants received training and technical assistance, and could bring challenges involving data to the hub, where possible, or use synthetic data. Governments, Pledger said, walk away with a prototype which they can build on or share with a vendor to build a solution based on already created specs.
Notably, the hub leverages an open source model so that if one government builds a solution, another can use it or build upon it for their own needs, Silverman emphasized.
Following the cohort, Pledger said she plans to apply the lessons learned within the city government.
“I think this is helping us as a city to understand where and how and when we could use AI,” she said, noting that her work in the cohort provided a “blueprint” that can be used.
