Convo Memory is a feature built into Convo that learns from your calls automatically. No setup, no extra software. After each meeting, it updates what it knows about the people you spoke with so the next call can pick up where the last one left off. It works best on one-on-ones and small group calls. Larger meetings are intentionally skipped.
What you need to enable it
Convo Memory needs to know who you are talking to. For that, it relies on your connected calendar. When a calendar invite says "you and Dmitry", Convo can attribute that conversation to Dmitry and start building a memory of him.
Without a connected calendar, Convo Memory has no way to map a recording to a specific person. Recordings still save and transcribe like normal, but no contact memory is built.
How to connect a calendar: Open Settings, go to Integrations, and connect Google Calendar or Microsoft Outlook. Convo only reads event metadata (titles, attendees, times). It never reads the meeting content of events you did not record.
What gets remembered
Convo Memory builds context from calendar meetings with four people or fewer (including you). A known contact is anyone who has been an attendee on a calendar meeting you recorded with Convo. For each known contact, Convo learns:
- What they keep returning to across calls
- What is still open between you
- What has been decided in past conversations
- Questions they raised that have not been answered
This surfaces in pre-call briefings, the Ask AI prompt during a live call, and the contact's detail view.
What gets skipped
For calendar meetings with five or more attendees, Convo Memory does not update any contact's summary from that meeting.
The reason is simple. Convo's transcription tells "You" from "Participant" but does not yet identify which specific person is speaking when several others are on the call. In a five-person discussion, attributing a comment to a specific attendee would be a guess. We would rather skip the meeting than learn the wrong thing about someone.
The meeting is still recorded and saved like any other call. You can find it in your call history, search it, and review the transcript. Convo Memory simply does not factor it into any contact's running summary.
What this looks like in practice
One-on-ones: full memory.
Three or four-person calls: full memory.
Five or more attendees: meeting saved, no contact memory updates.
Company-wide standups, all-hands: meeting saved, no contact memory updates.
A note on calendar invites
The four-person cap counts invitees on the calendar event, not who actually showed up. If you book a kickoff with six people invited but only three attend, Convo Memory will treat it as a five-or-more meeting and skip the contact memory update.
If a meeting consistently has more invitees than attendees and you want it factored in, the workaround is to remove uninvolved attendees from the calendar event before the call.
Walk-up calls and missed auto-detection
When you start a recording during a calendar event window, Convo links the recording to that event automatically and the memory follows. But what about spontaneous calls, walk-ins, or recordings where Convo missed the auto-detection?
From any conversation in your dashboard, you can manually link it to a calendar event after the fact. Convo Memory will pick that up and ingest the conversation against the attendees of that event. Same goes the other way. If you unlink a conversation from an event, Convo Memory will quietly forget that meeting from those contacts.
This means you can fix any miss without losing the context. A recording you forgot to schedule, a back-to-back call where Convo got confused. Pick the right calendar event and the memory rebuilds.
When this will change
Speaker-level transcription is on our roadmap. Once Convo can reliably attribute statements to specific attendees in larger meetings, the cap will be lifted. We do not have a timeline yet. This article will be updated when the work is in flight. For now, the cap is what keeps your contact memory accurate rather than approximately right.
Still need help? Ask AI about this topic