Defense and intelligence agencies play a critical role in supporting the military by detecting threats before they cause real harm.
This approach is not limited to the military. National security, cybersecurity, and law enforcement agencies are similarly leveraging intelligence gathering, surveillance, and enemy behavior analysis to focus resources on the “left side of the boom” in preparation for potential disruption or catastrophic events.
But for agencies like the Defense Intelligence Agency (DIA), these efforts are becoming increasingly difficult. One reason for this is the pressure to manage and process ever-increasing amounts of global intelligence data. Another is the need to keep up and adapt to the unprecedented pace and scale of technological change, DIA Chief Information Officer EP Matthew said at the recent GDIT Future Battlespace Summit hosted by Scoop News Group.
During a panel discussion with GDIT Senior Vice President of Intelligence and Homeland Security Aaron Bedrowski, Matthew noted that ChatGPT reached 50 million users in just five days, highlighting the accelerating pace of technology adoption. YouTube also achieved the same milestone 10 months. By comparison, radio and television took decades. Similarly, the compression of Moore’s Law (the time it traditionally takes to double the capacity of a semiconductor while halving the cost of a computer) has been reduced from approximately every two years to seven months.
At the heart of DIA’s challenge is the extent to which the government’s traditional five-year budgeting and planning cycles and acquisition schedules have not kept pace with corporate hardware and software upgrades. Matthew pointed out that semiconductor chips could improve in functionality by as much as 1,000 times in five years. This makes it difficult for government agencies to plan, acquire and deploy the latest technology, increasing the risk of national security vulnerabilities, he explained.
A modular component platform that is key to AI
Matthew argued that to survive technology’s accelerating innovation cycles, defense leaders must undertake a comprehensive cultural review and move from an infrastructure-centric, application-centric framework to an agile, secure, data-centric environment that can continuously deploy commercial software upgrades.
“Our goal is to create a data environment that focuses on accessing data only based on policy,” Matthew explained. Rooted in the Department of Defense’s Zero Trust cybersecurity framework, access to information is restricted by default. All ingested datasets are meticulously tagged, cataloged, and encrypted. Users are dynamically validated and granted access to data based on centrally managed policies. Importantly, even if data leaves the core repository, this security envelope remains in place, and user credentials are periodically retested to ensure absolute privacy and operational integrity.
This structured, policy-driven data layer also serves as an essential technical foundation for AI and advanced automated computing, he said, adding that organizations “can’t do AI unless they do the data right.”
GDIT’s Bedrowski noted that another challenge facing government agencies is how to integrate emerging technologies to scale faster. New features “happen so quickly that by the time we start integrating, there’s something new out there,” he said.
To extend AI and emerging technologies to existing intelligence operations and enhance decision-making, DIA is following a three-step process, Matthew said.
- First, data must be structured and constrained through fine-grained, centrally managed policy entitlements.
- Second, agencies are implementing modular component platforms (MCPs) to allow different components to capture data simultaneously.
- Finally, DIA networks must integrate semantic AI capabilities such as knowledge graphs, entity (data level) resolution, and graph (structure) resolution to turn raw, disparate information streams into meaningful operational intelligence.
By engineering the system using this highly modular blueprint, DIA expects to gain the structural agility to quickly replace individual software components as better commercial tools emerge. This approach prevents defense networks from being locked into long-term proprietary vendor contracts. It also enables a more seamless evolution from basic decision support to more comprehensive decision enhancement and automation.
Overcoming labor shortages through immersion
Ultimately, configuring and deploying these highly modular architectures requires specialized technology leaders who understand how to align new technologies to complex national security missions. Matthew acknowledged that the recent Defense Transformation Program (DRP) relocation has resulted in the retirement of approximately 22% of DIA’s network and software engineering professionals, significantly hampering the ability to build in-house expertise.
To combat this talent drain, DIA began eschewing the traditional post-sales vendor support model, which Matthew characterized as woefully inadequate for long-term operational success. Rather than purchasing commercial software or relying on outside contractors for maintenance, the agency established a hands-on internal training lab and launched an innovative “training with industry” initiative.
Borrowing from a similar Army initiative, this intensive program integrates DIA personnel directly into commercial technology companies and rotates them for six months. By gaining first-hand experience with cutting-edge software platforms, these operators will return to the agency better equipped for roles as in-house product leaders who can maximize existing software investments, Matthew asserted.
Additionally, the agency began engaging industry earlier this year to secure new capabilities to test, evaluate, validate, and validate new AI technologies.
This article was produced and sponsored by Scoop News Group for DefenseScoop. GDIT.
