With ZDR enabled, the storage step of the normal call pipeline is skipped entirely — recordings, transcripts, messages, and structured outputs are simply never written to Sulus's storage.
This does not change how the call is handled in real time. Orchestration-layer processing — the voice pipeline's endpointing, turn detection, and interruption handling described elsewhere in these docs — is always ephemeral by nature, regardless of whether ZDR is enabled. ZDR specifically governs what happens to content after the call, not the in-call processing itself.
ZDR is an organization-wide setting — it applies across your entire organization once active, it is not configurable per assistant or per call.