DHS, DOD Officials: AI is a Teammate, Not a Tool

Federal officials are emphasizing artificial intelligence’s place as one that augments the workforce to innovate and operate more efficiently.

“We’re entering an era where AI is not a tool, but a teammate,” said Defense Department Chief Digital and AI Office CTO William Streilein at the 2024 Adobe Government Forum last week. “If we embrace that, it empowers the user to think of AI, now that we can converse with our data, as a teammate. Understanding that aspect, I think is going to change the face of a lot of what we will be engaging in our roles.”

Streilein highlighted AI use cases for areas like intelligence analysis, where personnel can use generative AI to make probabilistic assessments of information and to simplify back-office functions to increase productivity across the board.

He pointed to Task Force Lima, which he said has provided guidance on and helped accelerate generative AI in the DOD.

Within the Department of Homeland Security, CAIO and CIO Eric Hysen said the agency integrating artificial intelligence and machine learning into its processes at borders has led to 200 successful examples of agents intercepting illegal narcotics like fentanyl.

“We’ve been working to refine machine-learning models to better look at data on passenger volumes of people coming in and out of the country to help our officers target and make smarter decisions about where they’re going to use those limited inspection resources,” Hysen said at the Adobe event. “With machine-learning models, we are now able to look at patterns in data, look at unusual patterns and give the officer information to still make the decision themselves. It’s always a human call on routing a vehicle for further inspection.”

Both leaders pointed to the importance of the private sector in building AI tools for government. Hysen noted that DHS is exploring pilots through open and closed source models with different providers to examine how AI can help the agency. He said DHS is unlikely to select a single large language model that it will use for the next 10 years, but wants to work with industry partners to develop the technology.

At DOD, Streilein cited the importance of partnerships to developing future technologies even beyond AI, he said.

“We’re working specifically to help us optimize our processes, optimize the policies, that enable us to experiment with the technology, leading to prototypes that can show operational impact, and then turn it into programs that continually get sustained and supported and perhaps even benefit from predictive maintenance,” Streilein said.

Streilein also emphasized the need to be cautious around AI development and that agencies should not cede control or responsibility of an algorithm. He said tools like artificial intelligence can help sort through data, but in the end “you are accountable as the human making the decision.”

Hysen echoed the sentiment of developing AI responsibly.

“All of our government organizations — agencies — make decisions that affect people’s lives in profound and significant ways,” Hysen said. “When we are thinking about any new technology, or any changes to how we think about those decisions, we have incredible responsibility to do so in a responsible, transparent and reliable way.”