Pentagon Officials Say Agentic AI Success Hinges on Data Integration

The military is moving beyond using artificial intelligence for data visualization and chatbots toward achieving “decision dominance,” according to War Department leaders.

Officials said during the AFCEA Washington, D.C. Chapter luncheon Wednesday they are shifting toward agentic AI designed to outpace adversaries by streamlining complex workflows and accelerating operational decision making.

“We found out that for us, the speed is really not [defined by] best algorithm, but the ease of which you can integrate the agentic AI into your workflow,” Army Chief Data and Analytics Officer David Markowitz said during the event.

Lowering the Barriers to Entry

Markowitz said that operationalizing AI requires dismantling bureaucratic hurdles that prevent rapid deployment. The Army is focused on “lowering the barriers across the scene” by reforming software acquisition and deploying “low-code/no-code” platforms that allow units to build tools in secure, local workspaces.

He pointed to the Army Data Operation Center, established in early 2026, which acts as a centralized support hub for divisions by ensuring authoritative data is readily available for rapid application development.

“It’s really starting to gel as we start to field next generation [command and control] from our two experimentation divisions. [It’s] starting to come back now to a central rear hub,” Markowitz said.

By prioritizing integration over purely technical superiority, Markowitz said the Army is building an ecosystem where agentic AI can be integrated into existing operations without requiring a specialized “ivory tower” of experts.

Col. Richard Becker, G6, Army Intelligence Security Command, said during the luncheon that agentic AI is useless without high-quality, accessible data. Becker stressed that the military must stop treating AI as a standalone technology and start treating it as part of workflow.

“When we integrate AI, it’s not about the AI. What we find really fast is it’s about the data,” Becker said “I want to make a t-shirt that says ‘It’s the data stupid.’ We just don’t have a good grasp from a department level … on where our data is.”

Data management remains the military’s most significant hurdle, Becker said, adding that the Army needs to push for “pure data centricity” — a goal to move away from stovepiped networks separated by classification levels. He said that the network should be a “commodity” that allows data to sit together securely, rather than a barrier that creates friction for the warfighter.

“We have got to get out of this [network-centric mindset]. The network is a commodity that gets the data where it needs to go. Right now, we have policies that direct us to have separate networks for different classifications and different partners and all of that,” Becker said. “And if we get to true data centricity, we need one network. Let all the data sit together. People and industry are doing it. Banks do it.”

Demystifying the Machine for the Workforce

Officials said that Pentagon components are grappling with a multi-generational workforce that ranges from “AI natives” to staff members hired decades ago. Agentic AI systems can be daunting for people who have never used them, Becker said.

“[I took an AI 101 class and] didn’t learn anything, but I’ve already practitioned. But for the guy that we hired 30 years ago to splice copper cable? He had a lot of questions,” Becker said. “[We need to be] demystifying what AI is, what it can do [and] what it can’t do. I try to remind people that an AI agent without any context or data is like a three-year-old kid. You’ve got to be very specific about what you tell it to do.”

To address education deficits, Jesus Caban, chief data and analytics officer at the Defense Health Agency, said his agency has implemented data literacy programs, including a “red teaming” exercise where 200 providers attempted to break AI systems to identify linguistic confusion and ethical risks. He said that education, training and guidance are the only ways to prevent system failure.

“I can send them out [to use agentic AI]. If people don’t understand the meaning how to use it, it will fail,” Caban warned. “[We need to focus on] AI, data, literacy, cloud, the technology pieces and workflows. If you’ve got a crappy process and you just automate it [with AI], you still have a crappy process.”