await transcribeMeeting(audio, { actionItems: true })

Lisa manages a 20-person remote team across 3 time zones. Every week, she sat through 8 meetings and spent more than 3 hours afterward writing up notes and action items by hand. She'd end up with a wall-of-text doc that half the team never read — and the other half would dispute which version of an action item was 'official.' Follow-through rates hovered around 40%, not because people were lazy, but because the assignments were buried in prose and lacked clear owners. After deploying OpenClaw with Whisper transcription on a $6 Hetzner VPS, her post-meeting workflow became: drop the recording file (MP3 or WAV from Zoom or Google Meet) into a shared folder. OpenClaw picks it up automatically, runs local Whisper transcription with speaker diarization, then uses a language model to extract a structured note. Within 60 seconds of the file arriving, the Slack channel receives a formatted summary with three sections: Key Decisions (for the record), Action Items (each with an @owner and due date), and Discussion Points (searchable reference). Lisa estimates she saved 12 hours per month — and more importantly, action item completion rates jumped from 40% to 85%. The difference: when an assignment shows up as '@bob: Draft API spec by March 10' in Slack, it gets done. When it's buried in paragraph 4 of a Google Doc, it gets forgotten.