An AI-powered public meeting-monitoring tool for Hearst reporters automates transcription, keyword detection and summarising of council and other public meetings.
In an INMA Ideas blog, editorial innovation and AI strategy vice president Tim O’Rourke tells how the Assembly tool harnesses the technology to expand local news coverage.
Even on the best days, he says, it had been impossible for the well-staffed San Francisco Chronicle newsroom to cover every Bay Area news event, “let alone the scores of public meetings that happen across the region every week”.
Assembly – Hearst’s AI-powered public meeting-monitoring tool – has changed that, and is also available to other group newsrooms including the Houston Chronicle, San Antonio Express-News, and (Albany) in New York, among others.
Developed by the Hearst DevHub team, it automates transcription, keyword detection and summarising of city council, school board, state legislature, and other public meetings. “It empowers journalists to efficiently monitor and report on events in their communities without being physically present or having to listen to the entire meeting,” he says.
“Reporters then delve deeper, making calls, verifying facts, and speaking with sources, which leads to stories and scoops they might not have uncovered otherwise.”
O’Rourke says the goal is to enhance the productivity and reach of Hearst newsrooms by automating the time-intensive process of covering public meetings. “Assembly addresses a critical need: enabling comprehensive monitoring of government meetings across our markets’ coverage areas while reducing the physical and logistical demands on reporters.”
Assembly’s robust technical foundation includes more than 200 custom web scrapers that watch government platforms hourly for new meetings. When a new meeting is detected, the recording is downloaded, the audio extracted, and the content transcribed using a self-hosted transcription system built by Hearst DevHub and powered by OpenAI’s largest open-source Whisper model.
The resulting transcript is uploaded to a Google sheet with hyperlinked timestamps, making it easy for reporters to locate key moments.
Reporters receive email alerts when a predefined keyword is detected in the transcript, and Slack messages provide a space for generating summaries or querying transcripts directly using OpenAI’s GPT-4o model.
Beyond summaries, Assembly also enables reporters to have a conversation with a meeting transcript, much like interacting with ChatGPT. “This lets them ask nuanced questions – such as what a council member said about a specific topic or whether a particular issue was discussed – and receive instant, contextual responses grounded in the transcript itself,” O’Rourke says.
“But all this work under the hood of Assembly is wasted if our newsrooms are not trained on how to use AI-powered tools wisely. To this end, every newsroom receives multiple training sessions on developing good habits around using Assembly as a tool.”
During these training sessions, reporters are reminded to always refer to the source material before publishing any takeaways from a meeting. They receive clear guidance on how to validate Assembly’s summaries via the hyperlinked timestamps in the transcripts that take them directly to the point in the meetings where details were discussed. “This ensures the tool enhances rather than replaces journalistic rigour.”
O’Rourke says Assembly has transformed how Hearst newsrooms monitor and analyse public meetings, allowing journalists to expand coverage without increasing their workload.
“One of our key objectives – building a scalable transcription pipeline – has been a cornerstone of this success. Between May 2024 and April 2025, Assembly transcribed 13,119 hours – more than 547 days’ worth – of government meetings from across the country. Journalists across our newsrooms have had Assembly generate over 1,500 summaries to get a rough idea of what was discussed before jumping into the transcripts.
“Perhaps most importantly, Assembly allows our metro markets to cover far-flung suburban areas they might not otherwise cover, making a dent in the proliferation of news deserts across the country.”
The success of Assembly has reinforced Hearst’s commitment to ethical AI practices and newsroom transformation. By streamlining workflows, fostering collaboration, and delivering cost-effective solutions, Assembly has positioned itself as an indispensable tool for local journalism.