Uploading meetings

Audience

Exec / consumer. This is the one place you’ll need the web UI.

Three ways in

Workforce0 accepts meetings three ways. Pick based on what you have:

You have…	Use this
An audio file (Zoom recording, mp3)	Upload a recording
A text transcript (Otter, Fireflies)	Paste a transcript
A live phone call to capture	Dial in — see Voice dial-in

All three end up in the same place: a Meeting object with a transcript attached, ready for the chief of staff to draft a brief.

Upload a recording

1. Meetings → Upload recording.
2. Drop the file. Supported: mp3, m4a, webm, wav, ogg, flac. Max 500 MB.
3. Pick the project this meeting belongs to (scope matters — see Projects & goals).
4. Optional but useful: add a title and a one-line context (“Quarterly review with design team”). The chief-of-staff uses this in the brief.
5. Click Upload & transcribe.

Whisper transcribes in the background. A 1-hour recording takes ~90 seconds on a modest server. You’ll see a status badge move from uploading → transcribing → ready. You don’t need to wait — the UI will notify you when the transcript is ready.

Domain-aware transcription

If you’ve built the project graph (see Project Graph), Workforce0 uses a domain-aware prompt during transcription. This means Whisper knows about your codebase’s class names, role names, and recurring nouns — so “BAAgentService” doesn’t come back as “BA agent service” or “be agent service.” The transcription quality difference on technical meetings is significant.

Paste a transcript

If you already have a transcript from Otter, Fireflies, Read, Fathom, or a manual notetaker:

1. Meetings → Paste transcript.
2. Paste the text. Plain text, Markdown, or speaker-labeled formats all work — Workforce0 normalizes them.
3. Pick the project.
4. Click Create meeting.

Pasted transcripts skip the transcription step entirely. The chief-of-staff picks up immediately.

Speaker labels help

If your transcript has speaker labels (Alex: …, Jamie: …), the chief-of-staff attributes decisions and action items correctly. Without labels it still works but loses that attribution.

Dial in

Give your meeting attendees a phone number. They call it, have the meeting, hang up — a transcript and brief land in Slack.

See Voice dial-in for the full setup. Gets you the lowest-friction capture: no recording, no transcript, no upload.

After upload: what happens next

For every new meeting:

1. The meeting becomes visible on the Meetings page.
2. The transcript is stored and indexed.
3. You (or anyone with approval rights) can Generate brief from it.
4. If the installer has turned on auto-brief, step 3 happens automatically the moment the transcript is ready.

The brief is then posted to your comms channel for review — see Approvals in Slack.

Privacy

• The recording file is deleted after transcription by default — configurable via the installer-side RETAIN_RECORDINGS env var (see Environment variables).
• Transcripts stay in your Postgres. They never leave your infra.
• If you use Whisper through OpenAI’s hosted API, the audio file transits OpenAI once during transcription — if that’s unacceptable, the installer can configure a self-hosted Whisper instead.

Editing or deleting a meeting

• Edit title / context — in place on the meeting detail page.
• Correct the transcript — click Edit transcript; changes are versioned so a bad edit is undoable.
• Delete — scoped by role. Deletes the meeting + transcript; does NOT retroactively pull briefs that were already generated.

What not to do

• Don’t paste personal / non-work conversations in. Workforce0 will dutifully draft a brief for your recipe idea; the AI quota cost is yours to own.
• Don’t re-upload the same recording if transcription is still in progress — it’ll queue a duplicate.
• If a meeting’s context is extremely sensitive (legal, exec compensation), ask your installer whether to route it to a dedicated locally-hosted model — see Local models (Ollama).