State and local governments face increasing -- and increasingly complex -- economic and social problems that require more agile decisionmaking with less certainty than ever before, as RAND describes in its Social and Economic Policy Rethink Initiative. However, there may be new solutions that can support government officials as they face the future.

AI technology is rapidly advancing, offering new capabilities for processing vast amounts of information and relief from administrative tasks burdening a shrinking public sector workforce. State and local policymakers lack systematic approaches to government AI integration. Efforts remain fragmented, leaving practitioners struggling with AI's evolving governance applications. In 2024, 150 state bills on the government use of AI were considered, 10 governors issued AI study orders, yet only 10 legislatures required comprehensive AI inventories. AI adoption carries risks requiring careful management and adaptive governance frameworks for these dynamic technologies.

Our recent article in Governing described the interested parties, approaches, government functions and potential benefits of AI-enabled governance (Figure 1). The purpose of this article is to describe what AI-enabled governance could look like in practice, particularly for dynamic and connected social and economic policy problems. We outline how these approaches can be implemented in a way that fairly maximizes participation and augments rather than replaces human judgment, keeping policymakers and the public in charge of final policy decisions. In short, this article begins to lay the foundation for a practical AI-enabled adaptive governance framework. We step through the complex case study of health, housing and the environment. These are areas that are being rapidly affected by AI usage in service delivery, but how should government use AI to keep pace on the policy response?

Severe housing shortages, rising greenhouse gas emissions, and growing health disparities linked with air quality and urban design are three interconnected crises that have been addressed separately, creating inefficiencies and missed opportunities. But let's consider how addressing these through an AI-enabled adaptive governance framework could help to maximize government benefits, using the framework domains above.

These are just a few examples of AI-enabled government functions that demonstrate the potential of AI to handle the complexity of implementing government initiatives at scale.

Now let's consider an AI-enabled adaptive government that performs all these functions simultaneously and where these functions interact with one another to create seamless real-time decisionmaking support for policymakers (Figure 2). For example, say a policymaker is considering a proposal to develop low-income housing units on the east side of a county and is trying to determine if these units are needed and if the proposed location is the right fit based on environmental risks and available services.

To help determine the suitability of the proposed low-income housing units, AI synthesizes local and state governments' housing, health, and environmental plans, budgets, and data on community conditions, and then matches community needs, local and state government services and programs, and housing, health, and climate policies to identify mismatches between housing needs, anticipated risks, and associated government efforts. This analysis suggests that low-income housing is needed in the county, but given the challenges facing communities on the east side, transitional housing is also needed to provide low-income families a path to more stable housing. AI also identifies the need for supportive services (employment, emotional well-being) and flood mitigation efforts to ensure these individuals can take advantage of the new low-income housing units. Based on the synthesis of community needs, AI also recommends additional "hot spots" in the county for placing new low-income housing units and transitional housing -- both in colocated and separated locations that would simultaneously maximize transit use, minimize emissions, and promote health. These recommendations are ranked by AI based on their alignment with forecasted future climate risks, population growth, economic development plans, and housing development.

AI then synthesizes these rankings for two audiences -- policymakers and the public -- and delivers the public synthesis to the local participatory governance platform. The public uses this platform to provide input, noting that the estimated housing need seems low and overlooks the large number of domestic workers in the community. AI uses this information to identify gaps in government plans and services and finds that local economic development plans don't account for these domestic workers and that school budgets don't account for the children of these domestic workers. AI then adds recommended "hot spots" that account for the public feedback and synthesizes this new information for the policymaker and the public. AI also recommends budget reallocations to support the services needed for transitional housing. The policymaker is able to quickly make a decision about the need and location for the proposed low-income housing units that is based on multiple data sources and public input. In short, AI allows the policymaker to make a decision that is informed by the potential cascading set of needs and conditions and best aligned with the goal of improving community conditions to optimize economic health and well-being (orange box under Community Conditions).

As state and local governments take up AI tools and technologies, there is no universal governance model, and there will be no one-size-fits-all adaptive governance model. Rather, AI governance must be tailored to state and local contexts and AI capacity (skilled workforce, funding, etc.) and include critical reflection on the outputs generated by AI tools. Additionally, there must be attention to the implementation of AI to ensure it is unbiased and accurate, and that sensitive and confidential information is protected, even as risks evolve. State and local governments may want to consider creating an Office of Algorithmic Accountability to ensure AI systems are responsible for their outputs or an AI Ethics and Oversight Bureau to ensure that the ethical and moral dimensions, as well as the practical monitoring of AI functions, include fact-checking of any AI-produced information. Additionally, policymakers may want to require an AI impact assessment to assess the potential human and financial damage of planned AI before its implementation. Transparency and trust in these processes will be essential to effective AI-enabled adaptive governance.

AI-enabled adaptive governance stands at a pivotal moment: it promises to revolutionize policymaking through seamless, real-time decisionmaking that responds quickly to emerging challenges, yet its transformative power makes thoughtful implementation not just important, but essential to determining whether this technology becomes democracy's greatest tool or its greatest vulnerability.