Every meeting now ends the same way: someone stops the recorder, drops the file into Speech Studio, and uses the transcript export to build the summary.
Total time: about three minutes of compute for an hour of audio, plus one minute to clean the output. That's it.
What changed internally: our notes are now searchable. The phrase we decided to use DuckDB for the launch report surfaces instantly when someone greps the archive. Before, that decision lived in someone's head until they left to join another team.
The one thing to watch: speaker diarization is hard. When three people talk over each other in a sprint review, the labels get fuzzy. That's normal — just review the section headers before you share the notes.
Try it on a one-hour recording first. You'll judge the transcript quality on your actual audio, not on demo files.