Meeting Intelligence
After a meeting recording is transcribed, Cadeo's AI generates a comprehensive meeting intelligence card with key insights.
What You Get
Summary
A concise 1-2 sentence overview of what the meeting covered.
Key Points
A bulleted list of the main topics discussed during the meeting.
Action Items
Tasks mentioned or assigned during the meeting. Each action item shows:
- What needs to be done
- Who it's assigned to
- A View task link if the action item was created as a Cadeo task
Action items work even when speakers aren't named. If the transcript has unnamed speakers (e.g. Speaker A, Speaker B), Cadeo still extracts every action item and assigns it using the exact speaker label from the transcript. This matters when meetings have external attendees, group-email invitees, or people who aren't in your Cadeo workspace.
Action items auto-refresh when you name speakers later. After a meeting, if you click a speaker label in the transcript and rename Speaker A to Josh, Cadeo automatically regenerates the summary and action items a few seconds later — so anywhere "Speaker A" was an assignee, you'll see "Josh" instead. Renaming several speakers in a row coalesces into a single regeneration to keep it fast.
If you'd rather refresh manually at any time, use the Regenerate Summary button on the meeting card.
Decisions
Explicit decisions made or conclusions reached during the meeting.
Where to Find It
Meeting intelligence appears as a card in your meeting note, with tabs:
- Summary — AI summary, key points, action items, decisions
- Transcript — full searchable transcript with speaker names and timestamps
- Notes — your manual notes on the meeting
Creating Tasks from Action Items
AI-detected action items can be turned into Cadeo tasks. You can also manually create tasks from meeting notes using the /task slash command.
Automatic speaker recognition
Cadeo learns to recognise voices over time. After your first few meetings where you manually name speakers, Cadeo builds a voice profile for each person.
In subsequent meetings:
- High-confidence matches are assigned automatically — you won't even see the speaker-matching card.
- Borderline matches show a quick inline prompt: "Is this Richard Crossman? ✓ ✕". One tap to confirm or dismiss.
- Unknown speakers fall through to the manual matcher card as before.
Voice fingerprints are extracted on your device and never leave your machine. The stored profile is a compact numerical summary (not a recording of the person's voice).
This feature requires the diarization models to be downloaded (Settings > Recording > Download speaker identification models). It works for both calendar-linked meetings and ad-hoc recordings.
Cross-Meeting Context
When viewing a meeting, past related meetings appear in a sidebar showing:
- Date and title of previous meetings
- Icon indicating if previous meetings have notes
- Click to jump to any previous meeting's details
This helps you build context before or during recurring meetings.