1. Home
  2. Help
  3. Workspace
  4. Meetings
Workspace

Meetings

Record any project meeting and get AI-extracted minutes — action items, decisions, and risks — each anchored to the exact audio moment it was said.

What you see

Meetings is the upstream surface where unstructured talk becomes structured commitments. Every project meeting — OAC, sub coord, owner check-in, site walk, safety stand-down, precon handover — is recorded, transcribed, and split into Action Items, Decisions, and Risks. Every extracted item is anchored to its [start, end] audio timestamp so you can play the exact moment it was said.

The three screens

List: every meeting in the project with type, status, attendees, and counts (N actions · M decisions · K risks). Filter strip + search across transcripts. Live recording: pulsing red bar, live waveform, optional markdown notes pane where you type what's important during the meeting. Finalized minutes: read-only doc with Action Items / Decisions / Risks / Carryover sections, each item shows a ▶ MM:SS chip that plays back the verbatim moment.

Every control

● Start recording

Begins live capture. Recording bar on top, audio waveform, your own markdown notes on the side. AI uses both signals at extraction.

↑ Import audio

Drop an .m4a / .mp3 / .wav from any recorder.

Type pill

OAC / Sub Coord / Owner / Site Walk / Safety / Precon. Drives default visibility (OAC + Owner default to team_plus_owner).

▶ MM:SS anchor chip

Click any audio anchor on a finalized item to play that exact moment.

+ Task

Promote an action item to the Tasks tracker. Bidirectional link stored both ways.

Carryover panel

Unresolved action items from prior meetings in the same series surface automatically on the next meeting's minutes.

↓ TXT / MD

Download the transcript + extracted items as plain text or markdown.

Finalize

Locks the meeting and triggers downstream wiring (Tasks creation, RFI resolution, Pulse indexing).

How extraction works

Audio runs through diarization and ASR (Gemini 2.5 Flash) to produce a time-stamped transcript with speaker labels. Then a second LLM pass extracts action items, decisions, and risks — using both the transcript and your operator notes. Items flagged in your notes get a confidence boost. Every output carries an anchor (start_sec, end_sec, verbatim quote) so the citation contract holds end-to-end.