How Hamlet turns opaque local government meetings into civic intelligence with AssemblyAI

City councils and planning commissions decide whether a warehouse gets built, whether a rezoning goes through, or whether a town installs license plate cameras. Decisions like these shape communities for decades. Most of them happen in meetings almost nobody watches. Hamlet is pioneering a new approach.

The company indexes city council, planning commission, and school board meetings across all 50 states and DC, then turns them into structured, searchable data. Contractors, site selectors, energy teams, and government affairs teams use Hamlet to see where a project may face pushback and understand who they need to talk to before a vote.

"The decisions made at the local level are very, very opaque and incredibly impactful to our day to day lives," says Sunil Rajaraman, Hamlet's founder. "Hamlet makes it clear what's happening at the local level."

Hamlet's products include timestamped transcript search, a daily intelligence brief that links every claim to its public-record source, and an API that returns data with sources.

Coming out of the COVID pandemic in 2020, most U.S. cities started recording their meetings. However, recording a meeting and recording it well are different things.

"Quality of equipment varies from jurisdiction to jurisdiction," Sunil says. "Agendas and procedures vary from jurisdiction to jurisdiction." Audio quality swings from city to city, accents and speaking styles change across the country, and meetings can run over five hours long, with council members, staff, and public commenters all taking the floor.

The team needed attribution to be successful. A transcript can tell you that a topic came up, but it doesn't tell you which council members opposed it.

"If you don't have good speaker identification, then you have no idea who views a certain issue a certain way," Sunil says. "That's why we can't use off-the-shelf YouTube transcriptions. We really need best-in-class transcription with speaker identification."

Hamlet started where most teams start: free YouTube captions. They weren't accurate enough, and didn't do a good job identifying who was speaking.

The team then worked through alternatives. They evaluated Whisper, built homegrown tooling on open-source models, and tested Gemini models, which struggled with the length of the recordings. Each option broke on one of three requirements: accuracy on uneven audio, speaker labels Hamlet could trust, and support for meetings that could run for hours.

"We looked at a few of these other tools," Sunil says. "None of them had consistent outputs and could handle longer meetings, which is what we're often dealing with on top of varying degrees of audio quality."

The decision came down to quality. "When you go to AssemblyAI, you're going for quality," he says.

The team at Hamlet runs meeting recordings through AssemblyAI's Pre-recorded Speech-to-Text API on Universal-3.5 Pro, with speaker diarization. The team is also starting to use Speaker Identification from AssemblyAI's speech understanding models to attach names to those speaker labels.

From there, their own pipeline double- and triple-checks who said what before a statement is attributed to a named official. Sunil sees this as the line between adding value and repackaging public records. Plenty of companies grab YouTube transcripts and charge for data that should be free. Hamlet wants customers to really understand where each of their elected officials stand.

"AssemblyAI adds real value on top of the data," Sunil says. "We want to get down to what Mayor so-and-so thinks of data centers or pickleball. In order to build that part of the product, you need very accurate transcription and very good speaker diarization."

Hamlet sells its data to companies that want to understand how a city views certain topics. On October 20, 2026, the team brought that data to every citizen.

Hamlet is launching a free tool for residents and journalists to look up their city council members and see their track records based on what they've said in public meetings. The launch covers 10,000+ council member profiles across thousands of cities. Each profile depends on attributing statements to the right person, every time.

"AssemblyAI helps us add 3x more value above what any foundation model could provide on speaker identification," Sunil says.

The team credits AssemblyAI's speaker accuracy as their edge over what a general-purpose model can do on its own. The team hopes that this promotes higher turnout and civic engagement around local elections.

City councils, planning commissions, and school boards meet every day, and Hamlet plans to cover more meeting types and more jurisdictions. "This is the largest source of original content in America," Sunil says. "As we expand meeting types, we want to scale our business with AssemblyAI."

Speaker accuracy is where the team expects the biggest gains. Today Hamlet still runs extra verification passes on speaker attribution. Every improvement in Speaker Identification cuts that work down. "As speaker identification becomes more accurate and viable, that makes my job a lot easier," he says.

The work has also taught him how much the country's speech varies. "There's all sorts of ways people speak, even within the United States," Sunil says. "How do you turn that into a common language? That's what AssemblyAI, in part, helps do."

Hamlet turns opaque local government meetings into data on where officials stand, for developers who need answers before a vote and for residents who want to know who represents them. That depends on transcription and speaker diarization that hold up on hours-long, inconsistently recorded audio from thousands of jurisdictions.

Sunil's advice to other teams: "If you're doing this at volume and scale, and you're dealing with disparate data sources that are just not uniform at all, AssemblyAI is a great solution. I would call it great for the big hairy monster problem of taking all these disparate sources and making sense of it."