NCIS, FBI officials see promise in artificial intelligence use cases for business side

As agencies continue exploring artificial intelligence tools, federal law enforcement officials cited successful applications of the technology in “back office” operations, including HR and cybersecurity.

During a fireside chat Tuesday at a AFCEA Bethesda event in Washington, D.C., Gregory Scovel, deputy director of operations for Naval Criminal Investigative Service, said that while the agency is “not yet ready to unleash an agentic AI tool into” its law enforcement data, it is deploying it for more administrative uses.

“We have had some success on the back office piece of that,” Scovel said. “And I think we’ll continue to do that into the future, while we work on the policy that is going to have to push forward in order for us to utilize those types of tools in sensitive law enforcement data.”

Scovel outlined two such back-office AI use cases at the agency: One is a tool to aid its HR processing and the other allows the agency to review the cybersecurity postures of vendors in its supply chain risk management work. 

Kevin Jones, a leader in the FBI’s IT Infrastructure Division who joined Scovel for the chat, similarly pointed to a tool the agency is working on to reduce the number of alerts cybersecurity workers receive that are false positives for malicious activity. 

He called the process of creating that tool a “good pathfinder” for deploying the technology and agreed with Scovel, saying that “not only on the investigative and operational side, but on more the business management of our big organizations, there’s a lot of opportunities for just efficiency in the triage and review that information.”

Jones also said the FBI is exploring the use of AI to help sort the tens of thousands of emails received internally on a monthly basis about potential phishing attempts. 

The examples highlighted a desire for use cases that are designed to aid workers with cumbersome business tasks rather than fully automating functions. The discussion, of course, also comes as the Trump administration pushes for efficiency in the government as well as major federal workforce reductions and agency restructuring that are poised to require the workers to do more with less.

HR, supply chain

According to Scovel, NCIS’s HR use case is aimed at speeding up its HR processes and has helped achieve that. 

“I’m happy to announce, NCIS can bring people on faster than anybody else in DOD right now, and that’s because of some of what that tool has done and the partnerships that we’ve built,” he said. 

The impetus for looking at an AI solution for HR was workforce optimization and shrinking back-office functions to try to get as many people as possible into the field to support its mission, Scovel said. But implementation didn’t start out smoothly. 

When the current “AI craze” began, Scovel said, NCIS created an internal version of ChatGPT to use on its unclassified system. That technology didn’t interact with any of its law enforcement data and was focused more on making its HR work more efficient. The catch? “It didn’t work. It didn’t work at all,” he said.

“Not only did it lead to a sense of frustration, but it forced us to kind of look outside and identify potential tools that we can use to make our business processes more efficient,” Scovel said. 

That’s when the agency brought in the current vendor, though Scovel didn’t name the specific company.

In response to an audience question about what NCIS learned from that initial failed deployment, Scovel pointed to issues caused by pushing out too quickly without a test group and a lack of communication to users. 

Since then, he said, NCIS has also set up a group to look at new technologies — regardless of whether they’re developed internally or externally — and test them before they’re rolled out. 

Meanwhile, its supply chain use case, Scovel said, has allowed the agency to point human intelligence analysts “in the directions where there [were] anomalies, and then utilize our investigative and operational authorities to go attack those anomalies to protect those supply chains.”

Scovel said the Navy spent almost $80 billion last year in research and development acquisition. For its part, NCIS is tasked with protecting that process “so when the war fighter presses the button, the machine does what it’s supposed to do,” he said. As a result, the supply chain use case was one of the agency’s first investments in AI.

While he didn’t name the exact tool being used, he said that the “nascent” use case is still being built out and NCIS is learning from it.

Cyber alerts

At the FBI, Jones said his team targeted the cybersecurity alerts use case because the “pre-canned” alerts workers were getting were overly broad and security operations center analysis — commonly known as SOC analysis  — were undertaking an “extremely manual” process to take two disparate data sets, compare them, and determine false positives.

Jones, who is acting deputy assistant director of the IT Infrastructure Division’s Infrastructure and Operations Branch, said the endeavor was a learning experience in the value of bringing all relevant groups into the conversation. 

“Just that exercise of being able to really define that use case was helpful, and I think, is something that we can apply, both within and beyond the AI space,” he said.

Exploring the use case also revealed some challenges for adoption, Jones said. It ran well in a test environment but he said they’re still working to get it into production amid workforce and funding issues.

“We’ve run into some issues with personnel cuts and funding, and so those are real challenges that I’m sure will continue in other use cases,” Jones said.