Cross-meeting memory

MeetCrew remembers meetings as structured records, not raw recordings — and it identifies people by platform identity, never by their voice.

What gets remembered

After a meeting ends, MeetCrew extracts structured records from the transcript: decisions, open questions, action items, notable statements, and the topics, people, and projects they involve, each carrying owners and dates. Memory is per deployment. Each teammate remembers only the meetings it attended, and one teammate’s memory is readable by another only when you explicitly configure that sharing.

Isolated and searchable

Every tenant’s memory is isolated from every other tenant’s — the separation is structural, not a filter you have to trust. Within a tenant, your teammates query memory through a search interface — structured search, natural-language search, and on-demand entity summaries — so a teammate can recall what was decided about an account or who owns an outstanding item.

No voiceprints

MeetCrew stores no voice-derived data of any kind. To recognize the same person across meetings, it uses platform identity — the Microsoft Entra identity, email, or display name — never a voice template. Speakers without platform identity are treated as anonymous, first-class participants. In v1, named speaker attribution is off by default and consent-gated.

Retention and deletion

Transcripts and structured memory are retained for about 24 months by default, tunable within a 30-day floor and a 36-month ceiling. Opt-out, a customer deletion request, or tenant deprovisioning removes memory across every tier. A “don’t remember” command prevents storage at ingestion rather than merely hiding it at read time.