Skip to content

Tenant Lifecycle Management

This chapter describes how a platform operator should create, configure, and maintain tenants in a multi-tenant deployment.

The goals are: - consistent onboarding (every tenant starts with a known baseline), - safe isolation, - predictable upgrades and change management.

  1. Provision
  2. Create the tenant container (organization record, identifiers, domain settings).
  3. Configure baseline
  4. Enable channels/integration endpoints as applicable.
  5. Configure transcription defaults (voice).
  6. Assign default permissions/roles.
  7. Attach baseline analytics configuration (fields + tasks).
  8. Validate
  9. Run ingestion + transcription + AI task smoke tests.
  10. Go live
  11. Enable production ingestion; monitor health and usage.
  12. Operate
  13. Support changes, add insights, handle overrides and governance.
  14. Deprovision
  15. Disable ingestion, stop processing, export if needed, apply retention and deletion.

A “baseline bundle” reduces time-to-value and support burden. Consider standardizing:

Baseline Custom Fields (examples)

  • CX: CSAT, sentiment, top issues, escalation reason
  • Conversation understanding: summary, topic, call reason, outcome
  • Sales: lead score, objections, competitors, next actions
  • Operational: language, channel type, disposition (if applicable)

Baseline AI Tasks (examples)

  • Conversation summarization (with explanation)
  • Sentiment classification (with explanation)
  • CSAT scoring (value + explanation)
  • Call reason/outcome detection (value + explanation)

Operator note: Prebuilt tasks can be published globally and default-disabled. Tenant admins can enable them when ready.

Tenant isolation and data boundaries

Document how the platform enforces isolation for: - conversation content (audio/transcripts/threads), - analytics outputs (custom fields, dashboards), - configuration (tenant tasks, overrides), - usage and billing metrics (if applicable).

Onboarding checklist (operator)

  • ✅ Tenant created and admin access granted
  • ✅ Ingestion configured (voice and/or text connectors)
  • ✅ Transcription validated (voice): transcripts exist for test calls
  • ✅ Global baseline AI Tasks available in tenant as Disabled
  • ✅ AI Assistant job processing confirmed for tenant
  • ✅ Dashboards/search available (if part of baseline)
  • ✅ Support and escalation contacts established
  1. Disable new ingestion
  2. Stop ongoing processing for the tenant (pause tasks/job eligibility)
  3. Export required data (optional)
  4. Apply retention/deletion policy (audio, transcripts, text threads, custom fields)
  5. Remove credentials and access (SSO, API keys, connectors)
  6. Confirm deletion completion (audit log)

Implementation notes

  • Global tasks appear in tenant's Disabled tab until the tenant admin enables them
  • Tasks with tenant-level overrides display an "Overridden settings" tag on the task card
  • Provide a "tenant template" mechanism with baseline fields/tasks (default disabled)
  • Track overrides explicitly and make them auditable (who changed what, when)
  • Treat deprovisioning as a first-class runbook with audit evidence