This interview was filmed on location at The Helix, Booz Allen’s Center for Innovation in Washington, D.C., as part of the event DE25: Driving Outcomes through Data. The program features top technology leaders from the public and private sectors sharing insights on cloud transformation, agentic AI, fraud prevention, and data governance. Through a series of dynamic conversations, the program captures how agencies are aligning digital infrastructure with mission needs to deliver real results for the American people. Watch the full show.
For more than a decade, the Central Intelligence Agency has led the federal government’s adoption of cloud computing, setting the pace for secure, scalable, and mission-driven technology transformation. In his segment on Driving Outcomes through Data, Larry Taxson—former Chief of CLOUDworks at the CIA and now Digital Capabilities Delivery Executive—provides an insider’s look at how the agency is evolving beyond foundational cloud infrastructure toward a sophisticated, AI-enabled, multi-cloud environment.
From global deployments to space-based platforms, Taxson illustrates how CIA’s cloud journey is no longer about migrating workloads, but about engineering strategic flexibility that keeps pace with the complexity of modern intelligence operations.
“We’ve been doing cloud for 13 years,” Taxson says. “But I believe we’re just getting started.”
Taxson reflects on the early days of CIA’s cloud strategy, beginning in 2013 with a single provider. That effort laid the groundwork for the agency’s foundational infrastructure, but the need for speed, adaptability, and advanced capabilities has driven CIA toward a new model—multi-cloud integration.
This new phase is about more than having multiple vendors. Taxson makes an important distinction between “multiple clouds” and “multi-cloud”: the former refers to running different workloads in separate clouds; the latter means architecting a single system that draws on multiple clouds simultaneously based on each one’s strengths.
“Let’s say one provider is better at language translation for Mandarin, another for Farsi,” he explains. “I want to access all of them in real time to meet our mission needs. That’s true multi-cloud.”
This strategic approach enables the agency to pull in the most advanced technologies from across industry and apply them to sensitive, high-priority missions—all within classified environments.
A key enabler of CIA’s modernization is the work of CLOUDworks, the internal team responsible for onboarding, accrediting, and operationalizing cloud service providers (CSPs) for use in classified environments. Taxson explains that CLOUDworks ensures secure access, billing setup, and integration into CIA’s global architecture.
With multiple CSPs now operationalized, the agency has greater flexibility to design modular, mission-ready systems that scale as needed and integrate cutting-edge tools such as artificial intelligence and managed services.
“Each of the CSPs brings something unique to the table,” Taxson says. “We’re working to make sure the right tools are matched to the right missions.”
CIA’s role as a global organization requires an architecture that can support operations everywhere—from dense urban centers to disconnected environments, and even beyond Earth’s atmosphere.
“We need to use compute, storage, and analytics at the edge,” he says. “And we’re looking at how to do that in space, too—because that’s where some of our mission takes us.”
While data centers remain an essential part of the architecture—especially for extremely sensitive data—Taxson sees hybrid models and localized processing as critical to mission agility.
Artificial intelligence is no longer a future ambition—it’s now a core requirement for the intelligence community. CIA has already begun integrating commercial large language models (LLMs) into classified systems, thanks to an accelerated push to bring modern AI services behind secure firewalls.
In partnership with CIA’s Office of Artificial Intelligence and the CSPs, the agency was able to go from requirement to operational deployment in under five months—a remarkable feat given the cybersecurity and performance demands.
Still, Taxson anticipates that the future of AI at CIA will be defined not by generic LLMs, but by small language models (SLMs) trained on bespoke data sets like imagery, signals intelligence, and domain-specific content.
“Commercial LLMs are powerful, but we need models tailored to the kinds of data we work with,” he explains. “SLMs will allow us to do that securely and effectively.”
One of the most important shifts Taxson outlines is a new way of thinking about industry-government collaboration. In the past, mission teams would relay requirements to integrators, who would then select technical solutions. But today’s rapid pace of innovation demands a more integrated approach.
Taxson proposes a “triangle” model: mission owners, integrators, and cloud providers must come together early and often to co-design systems that maximize capability, minimize risk, and accelerate time-to-field.
“We need our integrators to get smart on all the options out there,” he urges. “Don’t just wait for mission to tell you what they need—sometimes they don’t know what’s possible until you show them.”
CIA’s cloud journey is more than a technology upgrade—it’s a cultural shift. By building a flexible, secure, multi-cloud platform, the agency is creating a foundation for ongoing innovation in data processing, analysis, and AI deployment.
With classified environments now able to support the same pace of technology evolution as commercial settings, Taxson sees an opportunity not just to catch up with industry—but to help lead it.
“It’s about enabling the art of the possible,” he concludes. “And showing our mission owners what’s available—so they can dream bigger, move faster, and stay ahead.”