AI in Government Hinges on Supportive Leadership and a “Glass Breaker” in Charge

More technologically knowledgeable stakeholders and committed leadership within the federal government are critical to successfully implementing scalable artificial intelligence technology in public offices. 

Speaking during a virtual panel discussion, officials including Jack Shanahan, the inaugural director of the Department of Defense’s Joint Artificial Intelligence Center, discussed the need for modernization advocates in federal agencies, specifically needing a “disrupter” helming the implementation of new software. 

“If this is really the first big AI project designed to go fast and to go to scale, you need almost a classic glass breaker type of person that's just going to plow over all those bureaucratic obstacles,” he explained. 

He added that agency leaders and deputies need to be completely on board with new technology rollouts to help ensure that adequate oversight and accountability is present in artificial intelligence acquisitions. Shanahan also noted that more employees with a technical aptitude are vital. 

“You've got to bring someone in that is familiar with the way this happens in the commercial software world,” he said. Shanahan added that clear funding and investment is also a key component for agency artificial intelligence programs, namely Project Maven, an initiative to apply artificial intelligence to data collection via surveillance. 

Suzette Kent, the former federal chief information officer, agreed that the internal culture needs to respond to implementing more advanced technologies. 

“The more powerful the capabilities, the more skills…that have to be at the table,” she said. 

Artificial intelligence technologies that handle large amounts of data also pose challenges to efficiently scaling such tech. 

Deputy Director for Science and Technology at the Central Intelligence Agency Dawn Meyerriecks described her office’s experience with organizing data that is supposed to work with new machine learning technology.

“People really have to understand their data in order to make it available in a way that makes sense,” she said. Meyerriecks said she and her colleagues at the CIA ran several pilot projects to see what software worked best with different types of data collected by the agency. 

Shanahan added that the JAIC took similar steps in machine learning technology implementation, saying his organization focused on the steps within the AI stack, including forming a stronger pipeline between initial data collection and constructing a sustainable digital platform.

His team also worked to develop key infrastructure to support the initiative, including cloud, testing professionals, data protection strategy, centralized direction management and accreditation security. 

He further said that bringing these technologies into the government can help bridge gaps in public agencies introduced by legacy technologies.  

“There are all sorts of opportunities to be using AI and ML,” Shanahan said. “It's not gonna happen overnight, but you have to have sort of tech urgency. It's time to go beyond the small scale pilot projects and go to the fusion of innovation across the entire organizations. That's hard, but we don't have a choice right now.”