CDAO, the Pentagon’s AI-accelerating office, undergoing restructuring before presidential transition

In her final months as the Pentagon’s second permanent Chief Digital and Artificial Intelligence Officer, Dr. Radha Plumb and her team have been reshaping some of the hub’s directorates and acceleration cells to more quickly and strategically scale proven and experimental AI-enabled capabilities across the U.S. military at a pace that more closely matches real-world needs.

“The good news is it’s just a very natural evolution from what was already there,” Plumb told DefenseScoop Monday during an exclusive interview at the Pentagon.

When it first achieved full operational capability in 2022, the CDAO was structured around four combined predecessor organizations: the Joint Artificial Intelligence Center (JAIC), Defense Digital Service (DDS), Office of the Chief Data Officer, and the Advana program. Plumb, who before this role served as Deputy Under Secretary of Defense for Acquisition and Sustainment, assumed leadership after the office’s first chief, longtime commercial tech executive Craig Martell, departed in early 2024. 

After comprehensively reviewing the office’s inner workings, her team spent the last several months shuffling its structure to take a new path forward designed to expedite Defense Department components’ access to and adoption of AI for contemporary day-to-day operations. In the interview, Plumb provided DefenseScoop with a first look into the re-organization and the motivations behind it, as well as why she believes it makes sense ahead of her planned departure and the entrance of the second Trump administration.

“I will transition in mid-January, but the rest of the CDAO is career and technical expertise staff, and they will just stay and so a lot of the priority missions will continue,” Plumb noted.

Pressing needs

Putting it simply, some of the CDAO’s original teams — including those working on Advana, joint command and control pursuits, AI assurance, what was previously referred to as the algorithmic warfare group and others — have been renamed and reassembled into new efforts aimed at delivering in-demand AI and analytics across the enterprise, and via ongoing operational missions. 

“Over the last six months, it was really clear that two things were happening,” Plumb explained.

On the one hand, she spent time with CDAO colleagues focused on what she called “the integration that has to happen between those [DOD] customer needs, and the platform services” delivering products that meet them. 

Secondly, through steadily evolving efforts to propel the department’s realization of the military’s next-generation concept for combined all-domain command and control — including through its Global Information Dominance Experimentation series, better known as GIDE — Plumb said she and her team “really saw the importance of, early on, having that scaled capabilities view as we look at new solutions through these acceleration cells.”

Just as the CDAO is organized with lines of effort under policy and acquisition, for example, Plumb has established a new Scaled Capabilities directorate with a Senior Executive Service-level role and, for now, two existing divisions named “mission analytics” (MA) and “enterprise platform services” (EPS), focused on scaling capabilities to the enterprise.

While MA broadly includes CDAO officials working on customer support activities, the AI and Data Acceleration or ADA initiative and several others, EPS — formerly known as algorithmic warfare — is responsible for the underlying infrastructure, software tools and services for enterprise capabilities like the Advana platform.

Separately, but now reporting directly to the CDAO and operating in conjunction with that, is the existing “Advanced C2 Accelerator Cell” — as well as the new AI Rapid Capabilities Cell, or AI RCC, which Plumb unveiled last week. 

Describing the fresh vision, she explained: “When we see capability gaps at specific [combatant commands], how do we solve that pressing need — but then build that in a way that’s future-proof and can be scaled to other settings? So for instance, if we solve, as we have, operational real-time issues outside of the United States, then when we have major issues in the United States — like, say, a hurricane, and we need to optimize our operational response to that — how can we take those same tools and scale them to that new use case?”

The new AI RCC envelops maturing AI assurance and test and evaluation work, and CDAO-led efforts to facilitate the military’s responsible use of emerging and still-uncertain generative AI capabilities. 

Notably, the AI RCC (pronounced “arc”) marks the DOD’s next iteration of Task Force Lima. Plumb said it was in some parts born out of the learnings identified by that group. 

“I think one key finding was that there was a pressing need for the department to accelerate its identification of these AI capabilities, and then create pathways to scale,” she said, adding that the idea is to introduce “a small core team that works on pilots in the priority areas” paired with a team tasked with scaling such capabilities for wider use. 

In response to questions from DefenseScoop, she said the leadership team is “in the final stages with a candidate” who was recommended as a Defense Innovation Unit “pick” to steer the AI RCC. 

On Tuesday, a DOD spokesperson said Capt. M. Xavier Lugo — who led Task Force Lima through its duration — “has moved to lead another priority initiative at CDAO,” without clarifying which.

In terms of tangible generative AI progress the office has made this year, Plumb said that officials recently accelerated a large language model translation service identified by DIU for use across two military service partners. 

“We basically said, ‘OK, we know there are use cases. We know they’re [budgeting] for them in a few years — let’s not make people wait two years to have access to it.’ So we got a contract that allows the relevant customers to access it through, I think, Advana. But we’re working to integrate it on the Maven Smart System side too, so that our customer base can just start translating, leveraging, sort of the best of commercial tech out right now,” Plumb noted. 

The solution can take “a whole bunch of context and translate more like a person translator than a text translator,” she added.

Sometime in January, the CDAO also aims to open up new cloud-based sandbox testbeds for approved DOD users to experiment with different generative AI applications via the new cell. Plumb declined to share the cloud service providers involved at this time.

Crossover capabilities that enable C2 operations but function between the two rapid acceleration cells will be advanced in future GIDE experiments, according to the CDAO.

“The intent is that they should work themselves out of a business,” Plumb told DefenseScoop.

She pointed to a hand-drawn chart she produced during the interview to visualize the office and new moves, and used the Maven initiative as an example of the overarching approach.

“What we’re doing now is, we worked through identifying how we could scale Maven. We scaled it to a number of [combatant commands]. We’re expanding that scaling. So now, we’ve got the solution. Now that solution needs to move from this accelerator over here to our enterprise platforms, and I don’t have a timeline for you on that transition, but they’re working together — the Advanced C2 lead and our Scaled Capabilities and Enterprise Platforms lead to say: ‘OK what does managing this stack look like?’” she said. 

The CDAO confirmed that all of the new management changes aligned with the reshuffle are expected to take effect by Jan. 6. 

Change ‘during a baton pass’

Plumb acknowledged that upon taking over the AI office, she “put a lot of things on hold” to assess if the organization was operating at its best capacity — and areas where there’s room for improvement.

“I also pulled a lot of authorities up to me to create the management structures we needed,” she explained.

The CDAO’s new make-up comes as Plumb prepares to exit the office next month as a member of President Joe Biden’s departing administration.

“Transitions are a time of uncertainty and stress. I think the CDAO has really done a lot — has been through a lot, but it’s done a lot — to prove out how valuable this type of work is. This is a natural evolution in that process. That it comes at this time is just, we can’t lose time during a baton pass between administrations on these AI capabilities. And so we wanted to keep the momentum going,” she told DefenseScoop. 

Principal Deputy CDAO Margie Palmieri is set to serve as the acting chief of the office in the interim until Trump’s pick is named, and officials tapped to lead the reshuffled non-political positions are listed on the office’s recently updated website.

Plumb expressed confidence that steady bipartisan, bicameral support from Congress and commercial players — and the increasingly in-demand value proposition the CDAO offers to enable CJADC2 and more — will prove it even more valuable in the months and years ahead.

“I think the next three to four years are going to be probably dispositive on how the Department of Defense integrates AI into its warfighting and enterprise management. This is the time when the solutions are coming, and we know they can be transformational. So the next year, the team really needs to get those foundations, and they’re set to do it, get the foundation set, get those pilots going, and get the assessment criteria clear,” Plumb told DefenseScoop. 

“We’re almost fully staffed up now in terms of hiring personnel and had a big hiring push over the last six months. So we’ve got a team that is talented and has the vision and can execute. So, I look forward to seeing what happens,” she said.