Defense intel agencies build AI infrastructure before deploying agents

Defense intel agencies build AI infrastructure before deploying agents

U.S. intelligence agencies are preparing to move beyond chatbots to agentic AI systems, but their chief AI officers say they're prioritizing foundational infrastructure and risk management over speed to deployment.

Speaking at the DoDIIS summit in Tampa this week, agency leaders outlined how they're balancing the pressure to adopt evolving AI capabilities with the need for security, compliance, and operational readiness, according to Federal News Network.

Why it matters

The intelligence community's phased approach to agentic AI—systems that can take autonomous actions rather than simply respond to prompts—signals that even organizations under pressure to modernize recognize the risks of deploying advanced AI without proper guardrails. Their focus on data preparation and infrastructure could serve as a model for other federal agencies navigating similar adoption challenges.

DIA launches 90-day sprint for enterprise AI platform

Maj. Gen. Robert Kinney, chief AI officer at the Defense Intelligence Agency, said DIA is conducting a 90-day sprint to build its first enterprise AI platform. The agency is adopting Model Context Protocol, an open-source standard for connecting AI applications to external systems, and retooling its ChatDIA capability to support MCP and agents.

But Kinney emphasized the agency isn't rushing into agent deployment. "We're in the very early stages of thinking through that," he said. "You can't just jump and deploy into mission apps. You can't just jump and deploy into agents. You've got to build the underlying infrastructure first."

DIA plans to test agents with legacy data before advancing to mission use cases, building capabilities in stages while working through compliance, security, and zero trust requirements with IT teams.

NGA prioritizes AI-ready data

Michelle Aten, CAIO at the National Geospatial-Intelligence Agency, said her agency is focused on ensuring data is "AI ready" to avoid bottlenecks during rapid acquisition and deployment. "We don't want data to be the long pole in the tent," Aten said.

NGA has established a task force called ADROIT—AI and Data Return on Investment Tracking—to identify ROI for AI capabilities, benchmark performance indicators, and reduce redundancy across the intelligence community.

FBI manages 139 AI use cases through review board

The FBI is overseeing 139 AI use cases through an AI Review Board that evaluates applications not considered routine. Katie Noyes, the bureau's CAIO, said the FBI applies risk management on a case-by-case basis, following Office of Management and Budget guidance on high-risk AI.

"We're testing a lot more operational data and backstopping or paralleling with a human process," Noyes said. Some applications are producing strong enough outputs to warrant increased trust, while others require returning to development.

Rapid acquisition pathways accelerate AI adoption

Intelligence agencies are leveraging non-traditional acquisition methods to speed AI procurement. DIA has completed eight or nine other transaction agreements through its NeediPeDIA portal and a new consortium, reducing timelines from years to weeks or days, according to Kinney.

The FBI is using broad agency announcements and cooperative research and development agreements to connect with technology experts and evaluate commercial innovations before committing to contracts. "Now we can bring on that technology, we can benchmark and we can make an informed decision, and then follow that up with a lighter touch contracting pathway," Noyes said.

These details were first reported by Federal News Network.