Zoom screenshot. From Top left, Justen Fox of AJP, Ashley Harper, Product and technology officer at Charlottesville Tomorrow, Hilary Becker, Product director at The Beacon, and Matt DeFour, State Bureau Chief at Wisconsin Watch
Screenshot from the American Journalism Project's AI 2025 Showcase: Civic Information on Dec. 9, 2025. Clockwise from top left: AJP's Justen Fox, Charlottesville Tomorrow's Ashley Harper, Wisconsin Watch's Matt DeFour and The Beacon's Hilary Becker.

In December 2024, The Beacon joined 27 other nonprofit newsrooms in taking part in AJP’s Product and AI Studio.

We were exploring ways that AI could fit – or not – into our newsroom when we were accepted into this program. Our team was already working on an AI policy as part of our work with The Trust Project. People in our newsroom were dabbling with artificial intelligence on their own to help streamline some tasks or pull information from PDFs and images, and it was important to us that we approached this new technology with caution while keeping up with our industry.

We hit the ground running in January 2025. We set out with a big idea to leverage AI and large language models (LLMs) to transform public meeting transcripts and records into compelling story ideas for reporters to investigate further, with a focus on the Kansas and Missouri statehouses. Throughout the year we learned a lot about AI’s limits, pitfalls and uses. 

With this goal in mind we began working with our coach Justen Fox on what tools might be relevant to our projects. We got the opportunity to test Everlaw, ChatGPT Enterprise and LocalLens. What we quickly discovered was that many of these tools were not quite ready to follow statehouse committee meetings in the way we hoped. 

So we pivoted. And instead we asked: How can we use these tools to help in our newsroom? 

What we found was the LLMs were great at assisting reporters in smaller tasks for larger projects to help speed up reporting. For example, our rural health reporter Meg Cunningham used ChatGPT to parse through zip codes in rural Missouri to help find sources and locate potential attendees for our event about changes to the U.S. Postal Service. Our education reporter Maria Benevento was able to use LocalLens to look back and double check information from past school board meetings.

This cohort also allowed us to continue with a testing mindset. On the product side we created custom GPTs to test writing news quiz questions, design a health “brief” where our health reporters can ask for the latest health headlines across Missouri, and as a way as a way to find themes in large data sets 

We wrapped the cohort in December with three virtual showcases put together by AJP. Myself and Meg Cunningham were able to present our learnings at the civic information showcase hosted by Justen Fox, a coach in the AJP’s Product and AI Studio program. Fox actually put together a guide for newsrooms sharing the pros and cons of the LLM tools we tested – as well as ones tested by other newsrooms –  in the cohort.

At The Beacon, we aren’t looking to use AI as a tool to write our stories but as a way to enhance workflows. This cohort underscored that there are countless ways to harness AI for operations outside of writing high-quality journalism, and we look forward to continuing testing, questioning and innovating with these new LLM products.