On May 7, CFN stepped into a one-day hackathon hosted by our partners at Slingshot. Our role wasn’t to write code—we were the client, testing whether AI could really solve a pain point we see every day: neighbors can’t act on issues hidden in four-hour council videos and shrinking local-news columns.
We aim to empower more community members to stay informed, participate in policymaking, and strengthen neighborhood vitality.
Here is a summary of what happened but if you want all the details including an understanding of the full process and the technology tools that were used in the hackathon, the recording of the presentation with Slingshot that was hosted by the UofL Business School can be found here.
What we asked the sprint to prove
- Goal: Could a prototype digest the Metro Council meetings on ordinances through the process from Introduction through Passed, then return a plain-language brief within 24 hours?
- Why it matters: Dozens of neighborhood associations and advocacy groups now “reinvent the wheel” to track the same information. One shared pipeline could free them to focus on outreach instead of transcription.
What happened in 8 hours
- In the first part of the day, the development team ingested a sample of meeting agendas, ordinance documents, meetings transcripts and the like into the AWS platform, figured out how to organize the data needed for the analysis, and chose the LLMs to be used, which we’ll call the” back end.” In the future, these data elements will be “scraped” daily from the web and run immediately through the prompt engine to produce results.
- At the same time, the product and UI/UX teams worked with Mikal Forbush and Carla Dearing to build a compelling “front end in an AI product called Loveable.
- In the second part of the day, both teams were writing and passing prompts back and forth for testing to make the back end produce the data and analysis needed in the front end.
- By day’s end, we got to see a full prototype of the end-to-end product that would normally take 6-8 weeks of design and development work.
Key insights from the client seat
- Dashboard + Deep-Dive are both essential
Residents need a quick look (“headline banner on a stock-ticker screen”) and a path to drill into the details they care about. - Human-in-the-loop builds trust
AI speed is powerful, but editorial oversight is non-negotiable for accuracy and local context. - Chat future-proofs the interface
A conversational layer will let users ask new questions—“What changed in District 5 zoning this month?”—without redesigning screens each time. - Scalability looks realistic
The same workflow could cover zoning dockets and public-safety reports with modest additional cost, giving us a clear growth path.
It’s always powerful in product development to have all of the disciplines working together — design, development, finance, etc. — but doing this in AI made everyone feel like “we had a shared brain.” It was fascinating to experience.
Where we go next
- Field-test the prototype with residents, leaders and experts that follow this information professionally.
- Refine the chat layer so first-time users can ask anything in plain English.
- Validate a sustainable business model that serves both professionals and passionate residents. Many civic information efforts like this run out of funding after a few years.
This sprint didn’t finish CivicPulse—it confirmed that the investment is worth making. Stay tuned as we translate these insights into a public beta that puts real-time policy insight in every neighbor’s pocket.