But CDO Council Chairman Ted Kaouk, also the Agriculture Department’s CDO, said much of the work in the past year has come from agencies building up their IT infrastructure and workforce.
“Fundamentally, it’s about raising the expectations of the kinds of questions that people are asking, and then building behind that the workforce and the technology and the tools that are needed to get there,” Kaouk said Monday during the General Services Administration’s impact summit on data analytics.
USDA working under this approach has stood over 500 dashboards across its mission areas to give agency executives and employees better access to information.
Those dashboards give supervisors at the Forest Service, for example, an integrated view of metrics on timber sales, and give scientists at the Food and Nutrition Service’s Office of Food Safety access to timely data on laboratory sampling. Meanwhile, a COVID-19 dashboard has helped agency executives make key decisions about reopening facilities.
Kaouk said USDA took an iterative, human-centered design approach to stand up these dashboards, bringing employees together with more than 30 agencies executives to brainstorm cross-cutting questions, and then standing up prototype dashboards to answer to questions.
This exercise, he added, also revealed where the agency had data quality issues, or needed to fill in gaps in its ata.
“In a lot of cases, the findings that we had created a shared opportunity — that we had the data, but it might not have been available to those owners, and that led us in a direction of standing up a platform that could help us answer those questions,” Kaouk said.
To oversee this work, USDA has established a data governance board and stood up assistant chief data officers in each mission area. Kaouk said USDA has also spent time training its workforce on data visualization and spent much of the past two years working with GSA’s Centers of Excellence to strengthen its data analytics.
VA stands up data platform to track vets’ prescription drugs
The Department of Veterans Affairs, meanwhile, has stood up an enterprise data analytics platform called Rockies, which enables evidence-based decision making, improves health outcomes for veterans and their families, and lowers costs for the VA and taxpayers.
Denise Kitts, the chief data technology officer at VA’s Office of Information & Technology, said the data platform gives that agency a big-picture view of how veterans seek out and interact with agency services.
“There’s moments that matter in the journey that involve lots of different services that VA is delivering, and even folks external to VA that are delivering to veterans. And so we really needed a platform that could help us integrate data from lots of different programs,” Kitts said.
The data platform includes a zone-organized data lake that exceeds 3 billion rows of data, which has proven useful, for example, in working with the Office of Mental Health to determine whether veterans were taking their prescribed medications.
“They struggled with being able to do that in a timely fashion, so they were really excited about what we were able to do with this new modern platform and provide this information — these insights out to the frontline, the people that are delivering care to our veterans,” Kitts said.
JAIC seeks to make more data accessible, secure
While agencies face unique challenges in maturing their data infrastructure, nearly all must put in the work to organize messy and unstructured data.
To overcome this challenge, Marcus Comiter, the chief of the Data Science & Artificial Intelligence Engineering Division of the Defense Department’s Joint AI Center, said the agency is developing acquisition vehicles focused on services that allow agencies to prepare their data for use in AI algorithms.
The JAIC has also stood up a governance council aimed at setting best practices for data challenges unique to the defense community, such as organizing a wealth of video data.
“It’s easy to kind of get into this gold rush mentality, we just want to grab all of the data and bring it into our infrastructure,” Comiter said, “But what this data governance council allows us to do is to make sure that we have things like where that data came from, how it was prepared, who is the authority and owner of the data, all those types of things really written down, vetted and institutionalized.”
Other projects for the JAIC governance board include protecting AI algorithms and the data that powers from getting corrupted by adversaries, and building AI algorithms that are transparent in their decision-making. Comiter said the agency is also looking to make data discoverable and reusable for multiple projects.
“After we do all that work, we want to make sure that we get the maximum visibility for that across the department, so that the next component has to address a similar issue, they don’t have to start with data preparation at Step Zero, they can leverage existing datasets,” he said.