Using Multimodal AI to Explore Code for America's New Report on State AI Readiness

The Government AI Landscape Assessment is the first nationwide analysis of how ready state governments are to deploy artificial intelligence in public services. I was happy to lead the field research on this initiative by Code for America in my role as a consultant with HumanServices.ai (HSAI). After its publishing, I decided to explore multimodal AI tools for further dissecting the report.

The first step of course was Google NotebookML.  If you aren't familiar with the tool, it is a personalized research assistant that comes built with pre-prompted tools to analyze source materials including files and websites. It can also serve the purpose of a study guide builder. I connected the Code for America report website and it built a number of useful tools:

• Key Themes and Insights - A top level summary of the research.  Sample:

  Government AI Landscape Assessment: Key Themes and Insights

  This briefing document summarizes the key themes, findings, and future outlook presented in the "Government AI Landscape Assessment" published by Code for America in July 2025. The assessment evaluates the readiness of U.S. state governments in responsibly adopting AI, focusing on opportunities, risks, and human-centered implementation.....

  • Leadership & Governance: This dimension assesses "organizational structure and leadership dedicated to AI initiatives," including the presence of a Chief AI Officer (CAIO) or equivalent, cross-agency AI advisory groups, and overall governance approaches.
  • AI Capacity Building: This evaluates "investments in developing AI literacy, skills, and expertise across its workforce," including formal training, partnerships with educational institutions and industry, and upskilling pathways.
  • Technical Infrastructure & Capabilities: This focuses on the "technical foundation necessary to support quality data and advanced AI implementations," examining data infrastructure, accessibility, computing resources, platforms, and vendor partnerships.

  States are categorized into four readiness levels:

  • Early: "Initial steps in AI adoption with foundational elements beginning to emerge."
  • Developing: "Core components in place with growing capabilities and emerging formalization."
  • Established: "Mature implementation with systematic approaches and demonstrated effectiveness."
  • Advanced: "Sophisticated capabilities with comprehensive frameworks and innovative approaches."

  It is noted that "States may find themselves at different maturity levels across different categories, reflecting their unique strengths and focus areas."

• Mind Map - The mind mapping feature builds a clickable concept map from the source material provided.  Sample:

  

   

• Podcast Format Audio Explainer - Probably the coolest feature of NotebookLM is its ability to take your content and build a podcast between two people discussing and explaining your content.  The lifelike audio sounds like a real podcast.  Google also introduced a new feature that allows you to interrupt the podcast and ask the audio avatars a question.  They stop their podcast topic, answer your question, and then go back the podcast.  Mind blown!  Here's the audio file for podcast.

NotebookLM now allows you to share Notebooks publicly.  Click here to see the Notebook on Code for America's State Government AI Assessment.

I wanted to take the podcast one step further by turning into video. I tried out a number of AI video generation tools (a blog post for another day), but none of them worked too well for building a video conversation between two people, until I came upon Jogg.ai which had a pre-built podcast video builder from audio file or from text.  My audio file was too large so I went back to NotebookLM and asked for a shorter podcast. With this smaller video in hand, I uploaded and chose my podcasters from a list of prebuilt people visualizations.  The results were pretty good: