The Government Accountability Office, under its data strategy, is upskilling its analyst workforce as part and looking to elevate the overall data literacy of its workforce.
GAO, which turns 100 years old this month, estimates every dollar Congress invests in its budget flags about $114 in potential cost savings. But in order to maintain this return on investment, the watchdog is looking at ways to make its workforce more familiar with data analytics and artificial intelligence tools.
Taka Ariga, GAO’s first chief data scientist and the director of its Innovation Lab, said GAO’s data strategy will ensure the agency has the skills necessary to conduct the “audits of tomorrow.”
Emerging challenges include flagging improper payments amid the trillions of dollars in COVID-19 aid approved by Congress, as well as GAO having the capability to audit another agency’s AI algorithms and oversee increasingly large sets of agency data.
Ariga said GAO senior executives support the data literacy push, and that the agency is hiring digital natives into the organization. That, he said, leaves the agency focused on improving the data literacy of its middle management.
“They’re not always incentivized to think about, ‘Well, how can I do my data analysis better?’ They’re incentivized by meeting budget, meeting quality and meeting time. So how do we make sure that we have that enablement built-in, so that a middle manager can develop the type of data literacy that will actually augment quality and maybe reduce time and, ergo, meet the budget?” Ariga said.
Ariaga said GAO isn’t looking to train its whole workforce to become data scientists under this upskilling initiative. Rather than have them pursue a specialist position like data scientist, he said the workforce would be better suited by having a broader understanding of data skills.
“The way that we describe it is training them into general contractors, so that they know how to evaluate the work of specialists — [and] extending that same analogy, treating data scientists as a specialist in the realm of like electricians or plumbers. General contractors don’t need to know how to do that work specifically, but [they] know how to supervise [it],” Ariga said.
Data literacy is just one of three components of GAO’s data strategy. The agency is also focused on governance, making sure the data is quality and complete, and maximizing its data science and analytics capability.
Ariga said employees appreciate the upskilling effort because it trains them to be more well-rounded as auditors.
“Our employees recognize this sort of a traditional way of doing things is no longer sufficient, in a world dominated by AI, dominated by cloud services. This is not necessarily a punitive kind of thing to say, ‘If you don’t, augment your data literacy, you’re going to be irrelevant.’ But we really try to incentivize them as, ‘Here’s an opportunity for you to grow your career, here’s an opportunity for you to have the tools and the know-how to tackle sort of these evolving challenges,’ Ariaga said.
The data strategy, he added, is also looking at pivoting GAO from being an audit agency into more of a “knowledge transfer agency.”
“We want to make sure that we have a workforce that is capable of scoping data science aspects into their day-to-day audit engagement, but also be able to interpret the output of those analytical exercises so that we can make informed recommendations, we can make informed policy suggestions, and really, ultimately to drive for better and more accountable programs across the federal government,” Ariga said.
GAO has stood up a sandbox environment and on-demand video coursework, but Ariga said there are still some skills gaps the agency needs to address. GAO’s data literacy curriculum is focused on several “learning paths.”
“We didn’t necessarily have a curated curriculum to say, if you’re an analyst focused on, let’s say, a physical infrastructure issue, or focusing on energy issues or focus on national defense issues, there are very different considerations relative to the type of analytical techniques that might be generally applicable to those domains. We didn’t want to treat all analysis to be the same,” Ariga said.