Velocity by Booz Allen

We’ve decided to embrace the opportunities that large language models present

Booz Allen team working at an engineering lab in Panama City, FL

the owners, curation, and quality of that data so it can efficiently and successfully train and interact with the models. With a sound data management strategy, it becomes possible to navigate key security, privacy, and regulatory requirements and avoid data exposure and data loss while still embracing these large language models. Ultimately, it’s important to keep in mind that the investment to build a large language model from the ground up is likely to be prohibitive. In many cases, organizations can rapidly adopt and build capabilities on top of open-source or software-as-a-service solutions. And again, that begins with having a strong data management strategy, with data protection built in. We have lived this firsthand with our IT organization’s migration to a business systems data lake, which was foundational to our journey to modern data management and data protection. As I look forward, it will be especially important to further mature these concepts as generative AI opportunities are introduced. Q The technology industry is quickly adopting generative AI as an integrated piece of as-a-service products, many of which are already in use across organizations. How should IT leaders continuously evaluate their portfolio of industry tools without curbing employee access to valuable innovation? A Brad: So many vendors now offer generative AI capabilities that it can be difficult to prioritize service delivery opportunities within an individual enterprise. For this reason, it’s important for organizations to invest and innovate based on what their users truly need and what their larger data problems and essential use cases actually are. They can then sequence efforts with industry partners to collaborate on prototyping, testing, and piloting potential tools. This prioritized approach with multiple pilots allows organizations to develop and quickly refine governance processes for these third-party solutions and respond to evolving risks or changing regulations. We’re focused on giving our employees access to AI tools wherever they are needed to fulfill their mission with the right guardrails, guidance, and support. Initial areas of focus for us will be using large language models, natural language processing, and information retrieval to enhance IT support with chatbots. Our organization is working with key industry partners to pilot as-a-service capabilities that include AI-enabled knowledge management solutions and accelerated IT ticket resolution—all with the goal of improving employee self-service and automating employee- focused transactions. As a key design principle, we are pursuing goals regarding the standardization and modularity of these technology stacks and are intent on leveraging the power of generative AI tools while ensuring that restricted, proprietary, and other entrusted information stays compartmentalized within trusted environments.

while, of course, being proactive in managing

the high risks they carry. Beyond the hype, my focus is on using generative AI with purpose, in alignment with business objectives, and with a full understanding of the security requirements to safeguard sensitive information.” Q As IT organizations think about how to proceed with the latest AI technologies, what advice would your team share? A Brad: We are following the concept of “think big but start small,” and we advise other enterprises to follow this approach as they take on emerging AI capabilities. Making use of sandbox environments to test new use cases in a safe, controlled setting and then pursuing pilot programs is a good example of this concept. Participating in industry information-sharing groups, starting vendor AI capability discussions, and building acceptable usage of available services into training are other examples. In my experience, some examples of this approach include stacking models to optimize resource management, accelerating performance by using AI for vulnerability scanning and remediation, harnessing AI- powered paired programming capabilities to enhance employee productivity, implementing conversational AI capabilities that connect into our broader IT ecosystem, and accelerating task execution to significantly reduce time to completion across the enterprise. Ultimately, starting small with AI technology will enable enterprises

to fail fast and understand the impact of technical decisions before they affect real-world processes.