Velocity by Booz Allen

The Future of Generative AI in Government Missions Generative AI is swiftly taking hold across diverse government missions, revolutionizing processes—from data classification to research support—and enabling agencies to traverse and analyze vast amounts of data. Some of the most visible examples of generative AI are built to help the average user interact with the Federal Government. For example, large language models are enhancing citizen- facing chatbots, improving search functions, and personalizing service delivery (on page 58, read more about customer experience). But the use cases span diverse mission sets. One compelling application in development is a ChatGPT-powered platform for wargaming. Wargames realistic and complex simulations of stakeholders, such as threat actors, regulators, journalists, customers, and partners. A traditional simulation of multiple stakeholders’ responses and their interactions is time consuming and resource intensive. Additionally, the need for subject matter experts to play different roles can be impractical and costly, hindering the scalability and frequency of exercises. To address these challenges, ChatGPT can be imbued with a long-term memory platform to craft intelligent crisis management scenarios. These personas can dynamically respond to inputs, ask questions, make decisions, and interact with other personas and participants during both the planning and execution phases of an exercise. This new platform can turn static, open source data into a dynamically personas capable of simulating various stakeholders involved in responsive persona capable of responding based on data and its assigned attributes. The project aims to increase efficiency and deliver a highly realistic crisis management exercise. Another spotlight use case is in the space of infrastructure management. Generative AI is helping to create digital twins of complex infrastructure systems so that agencies can proactively monitor, optimize, and predict maintenance needs, ensuring and what-if exercises often lack the resources required to create

the efficiency, resilience, and longevity of critical government infrastructure. In areas such as bridge engineering or aircraft design, generative AI can benefit engineers by providing access to real-time data and analytics in addition to faster prototyping and rapid design iterations. As generative AI technology evolves, it will foster an era of cognitive augmentation where AI and human intelligence work together to accomplish tasks of an unprecedented scale and complexity. We see it expanding far beyond its present applications to impact areas such as: ORGANIZATIONAL PRODUCTIVITY By automating real-time data analysis of security reports, policy documents, and financial forecasts, generative AI can significantly enhance productivity. It eliminates bottlenecks, expedites decision making, and frees up crucial human hours. DIGITAL ASSISTANTS Powered by generative AI, digital assistants can learn from historical data, understand context, and generate humanlike facilitate information flow within government agencies, provide real-time writing and proofreading assistance, summarize complex documents, and create meaningful content. text. These assistants can streamline communication,

POLICYMAKING Generative AI’s capacity to simulate and generate diverse scenarios based on existing data makes it a valuable policymaking aid. It can predict potential impacts and outcomes of proposed policies, facilitating proactive decision making and aiding policymakers in drafting effective and comprehensive policies. DISASTER MANAGEMENT Generative AI can create hyperrealistic simulations of natural disasters, including complex climate models, helping to improve planning, resource allocation, and response strategies to minimize the impact of emergency events. SMART CITIES AND ENGINEERING By simulating traffic patterns, energy consumption, and population behaviors, generative AI could provide critical insights for planning urban infrastructure and public services to make our cities more efficient, livable, and sustainable. IMMERSIVE TRAINING Integrating generative AI with augmented reality and virtual reality marks a revolution in training and education within government agencies, providing immersive, personalized learning experiences.


generative AI can help predict and neutralize threats in real time. By constantly learning and adapting to new threat vectors, generative AI will be a powerful ally in maintaining national cybersecurity. There is an understandable wave of excitement and enthusiasm surrounding how generative AI technology will fundamentally shape how we live our lives, conduct business, and interact with organizations and with one another. To expand the mission impact of generative AI, government agencies and enterprises will need to understand how to design, build, and implement the technology and the use cases best suited for it. As an industry, it will take a commitment to establish the norms, standards, and regulations necessary for long-term use of generative AI, but for now it remains crucial for organizations to invest in deepening their understanding of the technology and its capacity for both risk and possibility. Ted Edwards and Alison Smith are AI technologists who specialize in the application of generative AI, helping Booz Allen and its clients harness the technology for mission impact.


Tim Lortz , a generative AI expert at Databricks, shared his perspective with our team. “Organizations which enable and encourage their people to incorporate generative AI into their work will be best positioned for long-term success,” he says. He pointed out that we also need to remember the human element—most of the world has only been familiar with generative AI for a few months, and the early adopters have tended to be

engineering types. “As managers, analysts, and other generative AI consumers gain a better understanding of what generative AI can, and can’t or shouldn’t, do for them, the use cases they identify and pursue will shape immense changes in the workplace and for our expectations with technology in general. These use cases could be as simple as writing or coding assistants built into familiar software tools, or something much more complex.” We asked Tim a few specific questions about generative AI: What advice can you offer IT leaders about exploring these capabilities? Getting started with generative AI has become relatively easy with the release of hosted generative AI models like ChatGPT and open source libraries like Transformers and LangChain to run generative AI applications locally. But going beyond demos to production applications is not for the faint of heart. It’s important for leaders to navigate key challenges and questions going into any investment, such as: How are you going to use generative AI models in secure, sensitive environments? How will you integrate generative AI into a broader data governance strategy? How can you effectively stitch together services from disparate vendors into your legacy applications? What are you most interested to see as this technology matures? Addressing the black-box nature of large language models is a crucial research area. Developing techniques to explain and interpret the decisions made by these models will be vital for their adoption in critical services. Also, it will be important to continue evolving open source models to more closely rival the generative AI capabilities of proprietary models like those of OpenAI. The gap is closing, but the difference in response quality is still noticeable. Can reinforcement learning from human feedback, for example, become easy to adopt for more organizations? Lastly, I’m also interested in developments around efficient model inference. The past few months have yielded impressive results in terms of “small” models with competitive generative AI benchmark scores. Improvements to attention architectures and quantization methods have shown a lot of promise to make models require less computation and memory without loss of quality, and we can expect to see more evolution.


Generative AI is transforming government missions by aiding in data analysis, communication, policy simulation, disaster management, and more. It is positioned to enhance productivity and decision making by facilitating collaboration between AI and human intelligence.

The promise of generative AI is undeniable. But to use it responsibly, enterprises will have to consider ethical implications, bias, and potential security vulnerabilities—and then understand and implement mitigation strategies.

To expand the mission impact of generative AI, government agencies and enterprises will need to understand how to design, build, and implement the technology and the use cases best suited for it.




Powered by