AI Resilience Report forLegislators

Right now, AI is mostly augmenting legislators rather than replacing them — and that's an important difference. According to the National Conference of State Legislatures, in Spring 2025 NCSL surveyed legislative staff about their use of generative artificial intelligence tools for legislative work, with responses from staff in 35 states, Puerto Rico and the U.S ^[1]. Virgin Islands, and the trend is clear: in 2024, 20% of legislative staff said they used these tools, but by 2025 that number had surged to 44%, with NCSL's lead on technology issues suggesting actual use may be even higher.

The most common tools? ChatGPT was the most commonly used generative AI tool, followed by Microsoft Copilot, both used to write content like articles and emails, assist with research, and condense hundreds of pages of material.

Globally, governments are experimenting further. A Tech Policy Press investigation ^[2] reports that the UK used an in-house tool called Consult to sort more than 50,000 public consultation responses in about two hours, while Italy's Senate uses AI to cluster similar amendments, and Brazil's Chamber of Deputies expanded its "Ulysses" program to classify legislative material. The Council of State Governments has been hosting workshops to help lawmakers understand AI policy ^[3], recognizing that legislators themselves are now the policymakers for the technology they're starting to use.

Still, the most human parts of the job — like the task of speaking to students to inspire future leaders (only 2% automatable) — remain firmly in human hands. Mentorship, persuasion, and listening to constituents in person can't really be outsourced to a chatbot.

Adoption is happening faster than many expected, but unevenly. On the speed side, the tools are cheap and widely available — ChatGPT and Copilot cost far less than hiring extra staff, which matters because lawmakers face chronic staff shortages, increasing policy complexity, and limited time. Route Fifty's 2026 outlook ^[4] notes that 2026 is less about adopting AI and more about embedding it responsibly into existing government work, with transparency and governance becoming "non-negotiable" as AI moves into mission-critical functions.

What slows things down is just as important. Legislators are uniquely cautious because they write the rules everyone else follows. Brookings researchers found ^[5] that states have proposed hundreds of AI-related bills but relatively few have passed, partly because highly educated states tend to introduce more ambitious and technically demanding legislation that fragments stakeholders.

There are also real worries about democratic legitimacy: AI-generated submissions could create what one researcher described as a "legislative DDoS" that overwhelms genuine public engagement.

For young people considering this career, the good news is that the skills legislators most need — building trust, listening to neighbors, persuading colleagues, and inspiring the next generation — are exactly the things AI is worst at. AI will likely handle the paperwork, but the leadership stays human.