The Capacity to Put AI to Work

In a recent conversation with university students, I found myself talking about a question I keep coming back to in my work: what will it take for government to put AI to good use? I enjoyed hearing how they approached it. As we talked, I realized how much of my own answer came from my time in government, and from seeing what people can accomplish when they have the right help.

In my work with state and local government, I keep seeing a gap between what AI can do and what an agency has the time, skills, and support to do with it. Someone might have a good idea for using AI, but someone also has to understand the process, try a different approach, check whether it works, and help colleagues get comfortable with it. Usually, those people already have full jobs.

I think of capacity as a rate limiter. The technology can keep improving, but how much of that improvement reaches people depends on our ability to put it to work. In government, that could mean someone getting help sooner, understanding a confusing requirement, or speaking with a public servant who has more time for their situation. Those are the possibilities that make the capacity gap worth caring about.

I worry about what happens if that gap grows. People’s expectations change as they use better tools elsewhere in their lives. If government cannot keep up, I worry it becomes one more reason to lose confidence in its ability to solve problems.

So I find myself thinking about diffusion, the dispersion of technology through society. What determines the rate of adoption? What slows it down? And what are the catalysts that could help? With AI, the technology we’re trying to adopt keeps changing. An agency can make real progress while the technology moves further ahead. Getting access is a beginning; learning how to use it well is ongoing work.

That is where cross-sector collaboration gets me excited. When I worked in San Francisco’s mayor’s office, I started Civic Bridge. We brought people from companies and universities together with city staff to work on problems the city had identified. Teams had 16 weeks to develop something useful, whether that was a technology solution, a design, or a policy recommendation.

I loved seeing people who would not ordinarily work together become invested in the same problem. City staff understood the work, while outside partners brought different skills and fresh eyes. Together, they could try something that would have been harder for either group alone.

That experience still shapes how I think about capacity. Useful knowledge is spread across institutions, and people often have more to offer one another than they realize. Bringing them together around a shared problem can change what they are able to do. I’ve seen how energizing that can be, especially when a problem that felt stuck begins to move.

I see that as one possible catalyst for AI adoption. Government, academia, industry, and civil society each have something to contribute, and each has something to learn. The part that interests me is how those exchanges can help people develop the confidence and judgment to carry the work forward themselves.

Of course, faster adoption is only worthwhile if it helps. I caught myself on that point during the conversation. It is easy to become enthusiastic about what the technology can do and assume that using more of it is progress. I want to be able to explain what gets better for the person asking for help and the public servant trying to provide it. That is what makes closing the capacity gap worth the effort.

I don’t have all the answers. But the conversation brought back something I loved about Civic Bridge: seeing people with different skills become invested in a public problem and begin to work out what they could do together. Talking with the students reminded me how much I enjoy that part of the work, when a fresh perspective helps me see a familiar problem differently.

The gap between what AI makes possible and what people experience when they turn to government feels like a problem worth spending my time on. I’ve worked on both sides of that gap. Helping people close it is the work I want to do.

Written with drafting and editing assistance from Codex. The experiences and final editorial decisions are my own.

I work at OpenAI. The views expressed here are my own and do not necessarily reflect those of my employer.