DISA sets the table for better AI with data management
Steve Wallace, the director of emerging technology at DISA, said a new tool, called Concierge AI, will reduce the friction to the user to find and analyze data.
The Defense Information Systems Agency’s data strategy is less than two years old, but it’s already ripe for an update.
The next version, under development, will put an even sharper focus on data quality and advanced analytics to improve how DISA uses artificial intelligence and other tools.
Steve Wallace, the director of emerging technology and chief technology officer at DISA, said a new tool, called Concierge AI, embodies the agency’s plans for integrating data with AI today and in the future.
“How do we augment our staff, leveraging large language models, and bring the sheer amount of data that we have, whether it be unstructured documents or structured documents, and deliver that to the user in a user friendly sort of manner?” Wallace said in an interview with Federal News Network. “Almost like a chat bot that you’re seeing in many different places, but how do we make that specific around the DISA mission, sometimes focused on the back office features, but then also with an eye on how do we do defense cyber operations (DCO) and help an analyst better do their job to dissect an attack and what have we seen before, based on after action reports and that type of thing.”
The overarching goal is to reduce the friction to the user to find and analyze data to drive better decisions. And doing all it in a way that uses natural language to make it easier on the employees, Wallace said.
That idea of reducing friction, making data easier to understand and use is central to the update to the DISA data strategy.
DISA wrote in a LinkedIn post on March 13 that the agency has made progress around setting up data governance and a data architecture as part of its implementation plan.
“The evolution of quality data and advanced analytics within the DISA community is the sole focus of the chief data office and will empower the agency to harness AI technology and AI situational awareness, predictive intelligence and decision-making agility, thereby enhancing national security and organizational readiness,” DISA wrote.
The updated DISA Data Strategy for 2025-2027 will focus on these mechanisms as part of the modernization and maturation of the agency’s data efforts.
DISA’s lessons in using AI
Concierge AI is part of how DISA is demonstrating the modernization and maturation. Wallace said the initial pilot is giving security folks and users a level of comfort in using the LLMs in a government-cloud Impact Level 5 environment.
“Some of the lessons we’ve learned is really around how do we secure these [LLM] environments? The concept of these vector databases is generally new, how do we secure them? How do we make sure that we’re doing the right thing by the data that we’re ultimately storing,” Wallace said. “I think we’re going to learn a lot as well as we start to ingest a large document set, which we haven’t necessarily done yet in the lab. It’s been very small dribs and drabs, but I’ve been encouraged by what I’ve seen just with the limited amount of what we have been able to do. In the first half of this calendar year, we expect to have something out to the digital workforce to start experimenting with, and from there, we’ll gather information about the user’s experience, and then, potentially, make it go more wide scale.”
One big opportunity to take better advantage of LLMs and AI tools is for defensive cybersecurity actions. Wallace said some of DISA’s cyber analysts already are modeling an attack, decomposing an attack and understanding exactly what happened by applying AI and LLMs.
“Any way that we can augment them by taking the datasets and feeding them into some sort of model to provide some sort of output so that they can have a one stop shop, if you will, to understand the dynamics of things that we’re seeing is probably one of the biggest ones that that’s out there in front of us,” he said.
Aside from AI, Wallace is also focused on several other priorities, including quantum encryption and rolling out classified mobile devices.
DISA awarded an Other Transaction Agreement to Sandbox AQ to figure out how to build a quantum resistant infrastructure in 2023. The prototype under development is for quantum resistant cryptography public key infrastructure.
“We’re in the phase right now of doing some crypto discovery. The OTA has, I think, eight different deliverables. We’re approximately halfway through it right now,” Wallace said. “This is about an education for us, and how we’re equipping the workforce to actually understand how some of these things work and the differences and, and watching as all of this evolves a lot more to come.”
The classified mobile device effort is further along. Wallace said he expects DISA to start rolling out the next generation devices in the coming months.