WASHINGTON — Even as doubts mount about the real-world benefits of chatbots and the possibility of an artificial intelligence bubble, the president of General Dynamics Information Technology said she was optimistic about AI technology, not just as a product to sell her government customers, but as a tool to empower (not replace) her own employees.
GDIT is already using AI to help its personnel find new jobs within the company, Amy Gilliland said, and it’s eagerly exploring AI’s potential to write code and process proposals as well.
What about the recently released MIT study [PDF] that found 95 percent of companies surveyed, as of this June, have gotten “zero return” on their investments in generative AI? When asked about that study and others like it, Gilliland argued it was too early in the adoption process to apply a framework like “return on investment,” at least the way GDIT is going about it.
“It’s early days,” Gilliland told reporters during a media roundtable Tuesday. “Right now, the battle that we are waging is education: It is about understanding what AI can do for you.
“When you’re talking about ‘return on investment,’ my investment right now is really about tailoring Large Language Models to my environment and figuring out use cases and how to apply them. [My] real investment is training employees on how to use AI and the tools that we have there to do their jobs more efficiently and better. … We have 10 times the number of AI training courses taken in 2025 as opposed to last year.”
Gilliland said she’s found her employees are enthused about using AI to do the gruntwork of coding large projects. The hope is that LLMs can write long passages of routine code that today require a lot of labor but little creativity, allowing the humans to focus on more intellectually challenging and interesting work.
“They don’t want to eat the broccoli, they just want to have dessert,” she chuckled.
Gilliland also sees potential for LLMs to help draft proposals, especially in finding key points buried in tedious government solicitation documents and finding comparable past projects to build on. “We bid … over a thousand opportunities every year,” she said. Even with AI, “some human [will] probably touch every page of that,” but LLMs can at least do a first, fast pass and get the process started.
“I’m not replacing those employees,” Gilliland emphasized. “I am helping those employees say, ‘Okay, these are the things that I need to focus on.'”
Already, GDIT has implemented an AI tool for “internal mobility,” Gilliland said, which helps match employees to new opportunities within the company. The algorithm facilitates both permanent changes of position inside GDIT and short-term “gigs,” which are temporary assignments from “six days [to] six weeks” assisting a different GDIT project full or part-time.
NGA’s New Neighbor And The Need For Speed
In fact, the occasion for Gilliland’s Tuesday roundtable was the unveiling of a prime location for such short-term assignments: GDIT’s grandly named new “Mission Emerge Center” just outside DC.
The 5,000-square-foot facility is essentially a showroom and meeting space where GDIT execs and techies can demo new AI and other products for potential government customers. On Tuesday, reporter were shown AI analysis of surveillance video, 3D mapping software, and a chatbot meant to help non-experts use sophisticated geospatial intelligence tools.
GDIT says the Center runs all this software in unclassified “sandboxes,” unconnected to any operational systems. That allows government customers can play with it and offer feedback before they commit to the time-consuming, costly process of getting it vetted to run on their own highly secure networks. The Center is also located in Springfield, Va., less than five minutes’ drive from a major campus of the National Geospatial Intelligence Agency (NGA), which is both one of GDIT’s major customers and a voracious consumer of AI.
All told, Gilliland said, new technologies like AI and new venues for cooperation like the Springfield center should allow the military to upgrade its technology much faster, in weeks or days instead of months and years.
“Ukraine is teaching us something about that right now,” she said. “We’re watching pretty scrappy Ukrainian troops figure out how to use technology in new and different ways … in real time.”
So, nowadays, after one of her product teams briefs a military or intelligence customer, the response is no longer “come back in a year,” Gilliland said. “It’s like, ‘hey, I’d like to have you back here in six weeks from now, because I want to do Agile sprints and see how this is working and evolving.’… It is that pace the DoD says they want” — especially after Secretary Pete Hegseth’s call to arms for industry execs last month.
That’s a big change from traditional turnaround times. “On a ship, it used to be that when you came back into port from deployment, they would literally cut a hole in the side of the ship to take the last deployment of software out of the ship, and put the next deployment of encryption and software on,” said Gilliland, an Annapolis graduate and former Navy officer.
“You had to go 18, 27 months, however long the deployment period is, to get the next iteration. Well, I’d like to get that while I’m underway. That is where we’re headed, and I think that’s where we need to be,” she said.