State legislatures face a widening “pacing problem” as generative AI reshapes lawmaking faster than institutions can adapt. Drawing on interviews across twenty-two states, this report examines staffing, institutional knowledge, network relationships, and technology adoption, revealing where legislative capacity lags and where states are quietly outpacing Congress.
By Caitlin McNally, Anne Meeker, and Aubrey Wilson
American state legislatures have long served as the laboratories of democracy, but they have modernized unevenly. Generative AI is now testing that gap in real time. In January 2024, POPVOX Foundation's Representative Bodies in the AI Era: Volume 1 documented how Congress and parliaments abroad were beginning to adapt to AI, and flagged state-level adoption as a thread worth following.
This report does three things.
First, it assesses state legislative adoption against the approaches, use cases, and institutional timeline outlined in Representative Bodies in the AI Era: Volume 1. The central finding is that state legislatures are simultaneously ahead of and behind the trajectory Representative Bodies anticipated, depending on which dimension is measured. A handful of domain-constrained platforms, including Iowa's Legible and Arizona's Skywolf, have reached the medium-term integration Representative Bodies projected years out. At the same time, most legislatures have not completed the basic immediate-phase work of issuing use guidelines or funding experimentation, and the most advanced implementations documented in this research are frequently the product of individual legislators or staff members rather than institutional strategy. Where formal capacity is thin, adoption does not disappear; it becomes privatized, improvised, and unevenly distributed.
Second, it documents how legislators and staff are actually using AI across twenty-two states, creating an initial field-based inventory of state legislative AI adoption. The unevenness of AI adoption this report documents traces directly back to four capacity dimensions this report examines: people, institutional knowledge, the broader network of relationships extending past the legislature's formal borders, and technology readiness. Thin staffing leaves offices no bandwidth to evaluate tools, so Microsoft Copilot becomes many staffers' default first experience with AI and, by several accounts, their most discouraging one. Institutional knowledge in state legislatures still lives mostly in individual memory rather than documented systems, leaving little for AI-assisted work to draw on and nowhere for training to live. Legislatures depend heavily on the executive branch, peer states, and outside organizations to fill research and drafting needs the institution has not built internally, a dependency now repeating itself in AI governance. Finally, technology readiness — digitized records, procurement processes built for AI, and clear governing policy — determines the ceiling on what any tool can do, regardless of how powerful or well-designed it is.
Third, it examines what those adoption patterns reveal about legislative capacity itself and proposes additional dimensions for assessing whether representative institutions are equipped to adapt to rapid technological change. This report's findings indicate that traditional political science comparative legislative capacity metrics of salary, session length, and staff numbers are no longer adequate to assess legislatures' ability to adapt to changing conditions. Section 5 proposes additional dimensions for assessing legislative capacity as part of a new Legislative Adaptive Capacity Index (LACI), including demand capacity, knowledge infrastructure, network capacity, and technology readiness. While beyond the scope of this paper, the prospective framework may serve as a guide for what legislatures should build, not only a way to measure what they already have.
Section 6 carries this into thirteen recommendations, adapted from Representative Bodies and organized around three priorities:
The choices state legislatures make on these fronts over the next several sessions will shape whether AI expands their capacity to govern or deepens the asymmetries this report documents.
Representative Bodies anticipated the institutional requirements accurately, even as the technical means of meeting those requirements continue to evolve:
Representative Bodies Expectation |
What the State Research Found |
What Changed or Requires Refinement |
|---|---|---|
| AI would assist with legislative summarization | Members, counsel, and caucus staff routinely use it | Institutional verification and authoritative sourcing remain weak |
| AI would expand constituent engagement capacity | Offices are using agents for intake and topic analysis | AI is also manufacturing or amplifying constituent demand |
| Domain-specific institutional tools would emerge | Legible and Skywolf represent early examples | Technical architecture matters less than governance, evaluation, and access |
| Legislatures should issue agile guidance | Idaho and Washington show why guidance accelerates adoption | Most institutions remain behind their individual users |
| Human oversight would remain essential | Members are citing generated outputs as authority | “Human in the loop” must mean source-based review, not nominal responsibility |
The Pacing Problem and GenAI's Impact on Representative Bodies • What "Capacity" Means in This Report • The Vol. 1 AI Adoption Approaches and Phases Framework Applied to State Implementation • Roadmap
Drawing on original interviews across twenty-two states with legislators, staff, and civil society partners, this section details the practitioner-driven methodology behind the report.
Where State Legislatures Sit on the Representative Bodies Adoption Phases • Use-Case Inventory • What Adoption Is Already Producing: New Pressures and Risks
People: Pressures and Elastic Capacity • Institutional Knowledge • The Broader Network • Technology Adoption
Since traditional metrics like salary and staff size miss AI-readiness, a new Legislative Adaptive Capacity Index measures people, knowledge, network, and technology to guide investment.
Build the Institutional Foundation • Move the Adoption Curve • Govern What Adoption Produces
State legislatures show uneven AI adoption tied to institutional role more than geography, with gaps in people, knowledge, and infrastructure raising open questions about partisanship and civil society's capacity.
