How to use Claude to turn meetings into queryable notes

FireBeyond built a meeting pipeline to address transcription errors in colleague names, customer names and internal terminology. Krisp supplies the raw transcripts. The author hands each transcript to Claude with a manually prepared context document containing a people directory, terminology glossary and master transcription error table.

Claude follows a fixed sequence: read the context, read the transcript, and reconcile uncertain names or terms against the error table before drafting. It asks only about unresolved cases. After the author confirms those, it produces a Markdown summary with an overview, topical notes, decisions and action items. Work items are pushed into Todoist.

The summaries go into Obsidian with YAML frontmatter and live Dataview queries, making meeting metadata and open actions behave like database records. Through Cowork, the author queries the accumulated folder with questions that require reasoning across separate meetings, including whether earlier customer requests have led to shipped work.

Resolved clarifications are written back into the context document. The author says this means previously corrected errors are handled automatically on later transcripts, and summaries have become tighter with less intervention.

Who it's for: teams with recurring meetings and specialized names or terminology that transcription tools frequently misrecognize.

Prompt
Suggested prompt (written by Agents In Use, not the author's)

Source: Hacker News · FireBeyond