AI is becoming more accessible to the next generation of workers—those who grew up surrounded by smart devices, self-driving cars, and instant digital connectivity. For these “digital natives,” AI feels familiar and achievable, shaping their expectations about what’s possible.
This shift was a key theme at a recent panel discussion during AI World Government, focusing on the mindset and skill sets needed for AI teams working in government. The event, held both virtually and in person in Alexandria, Virginia, brought together experts to explore how government agencies can harness AI effectively and responsibly.
Dorothy Aronson, CIO and Chief Data Officer at the National Science Foundation, emphasized that while AI technology is readily available, our cultural understanding and maturity haven’t caught up. She likened giving powerful data tools to giving a sharp object to a child—you need to be careful about how you use such tools. As AI accelerates, expectations rise. Vivek Rao, a researcher at UC Berkeley, reflected on how tasks that once took months—like natural language processing research—can now be completed in days, thanks to immense computing power and the enthusiasm of students who see AI as a game-changer.
Rachel Dzombak, who leads digital transformation at Carnegie Mellon University’s Software Engineering Institute, asked what makes working on AI in government unique. Aronson pointed out that governments must strike a delicate balance: they can’t move too fast with new tech, or users may be left behind. Unlike consumer tech products like iPhones, government AI projects require experimentation and foresight, especially as they serve a diverse workforce spanning generations. She shared her own initial misconceptions about working in government, which she now sees as a mission-driven environment dedicated to solving big societal problems like equity, safety, and access to resources.
Vivek Rao shared how students are often surprised to learn that much disaster response work in places like California involves local, state, and federal agencies working together. He developed courses on innovation in disaster management, which have inspired students to consider careers in government technology—one of whom is now a software engineer with a defense research organization.
Aronson also highlighted the challenge of onboarding new federal employees, describing it as a “heavy lift” that could be streamlined with better preparation.
When it comes to building effective AI teams, Bryan Lane, Director of Data & AI at the General Services Administration (soon transitioning to the FDIC), stressed resilience. He explained that successful teams need to be prepared for surprises and setbacks, maintaining focus on their mission even when faced with the unknown. Lane also noted a positive trend: team members increasingly admit when they’re venturing into unfamiliar territory. This honesty fosters a culture of learning and growth. Aronson pointed out that convincing management to pursue AI projects can be tough, especially when outcomes are uncertain and costs are unpredictable.
Rao encouraged fostering an experimental mindset—recognizing that accessible AI tools can mask underlying challenges. He advises that when applying these tools, organizations should be prepared for bumps along the way.
In discussing team-building, Aronson advocates for diverse groups focused around specific problems rather than tools. Lane agreed, emphasizing that success depends less on the technology and more on understanding the core issues and bringing together the right people. He prefers forming cross-functional teams that work intensely for about 45 days, often involving internal customers who learn more about data management and AI along the way. These collaborative efforts can produce advocates who push AI initiatives forward across the organization.
Lane envisions a five-year horizon to refine best practices for government AI projects. He cites the US Census Bureau’s Opportunity Project, launched in 2016, as a prime example. Over the years, TOP has facilitated over 135 projects addressing issues from ocean pollution to pandemic recovery, engaging more than 1,300 participants—from developers and policy experts to students and government officials. This initiative is based on a proven approach to organizing work and scaling successful methods.
Looking ahead, the consensus is clear: building a resilient, adaptable, and diverse AI workforce in government requires patience, experimentation, and collaboration. As the landscape evolves, so too will the strategies for harnessing AI’s full potential to serve society effectively.
