The 90,000 different local governments in the USA provide essential services and protections for the American people, including running schools, libraries, and public parks, and delivering front-line services during emergencies. While governments strive for efficient use of time and resources, due to the variety of size, structure, and complexity of local governments, it is currently prohibitively difficult to collect data and study these functions comprehensively. As a result, much of the research into the effectiveness of service delivery focuses on institutions at the federal or state level. This project will improve the capacity for studying local governments service delivery by providing new datasets and tools that make it possible to do comprehensive, comparative research at scale. Doing so will empower local governments with new research at the intersection of public policy, civic engagement, and artificial intelligence. This proposal builds upon the team's current open-source testbed. The long-term vision for the testbed is to facilitate rigorous experimentation through the robust, secure, and safe deployments of AI. To achieve this vision the proposed planning grant will carry out four activities: 1. Prototype safely integrating AI in the current testbed; 2. Convene government staff and researchers across a variety of computational fields to build a research agenda; 3. Govern robust AI deployments to operate safely and manage risk; and, 4. Validate the ability of