Skip to content

Getting Started Checklist

This checklist helps you launch Conversation Analytics quickly and safely. It assumes you are a tenant administrator.

Phase 1 — Verify prerequisites (30–60 minutes)

  • Confirm you can access Custom Fields
    Administration > Customization > Custom Fields
  • Confirm you can access AI Tasks
    Administration > Speech Analytics > AI Assistant > AI Tasks
  • Confirm you have sample conversations with transcripts (for calls) or message threads (for text channels)
  • Confirm you have at least one internal stakeholder for validation (QA lead, CX leader, Sales manager)

Phase 2 — Enable transcription (voice calls)

  • Enable and validate transcription (see: Enable Transcription (Prerequisite))
  • Spot-check 10–20 recent calls for transcript quality and language correctness

If you do not see transcription settings in your tenant: transcription may be managed by MiaRec or your service provider. Use the validation checklist in the transcription chapter and contact your provider if changes are required.

Phase 3 — Enable “starter” AI Tasks

In Administration > Speech Analytics > AI Assistant > AI Tasks:

  • Open the Disabled tab
  • Enable these starter tasks (recommended):
  • Conversation Summarization
  • Sentiment Analysis
  • CSAT (if you want CX scoring right away)

You can start with only one task. Fewer tasks = easier validation and lower cost.

Phase 4 — Test and calibrate before broad rollout

  • Use Playground or Save and Test to validate each enabled task on 5–10 representative conversations
  • Confirm outputs are:
  • present (not blank)
  • sensible (value matches transcript)
  • consistent with your definitions
  • accompanied by a short explanation (where configured)

If results don’t match your expectations:

  • Override Prompt, Filter, or both (see: Override Prompt and Filters)
  • Re-test until you reach acceptable consistency

Phase 5 — Make insights actionable

  • Confirm insights appear in Conversation Details (Analytics)
  • Configure/verify dashboards for key metrics (CSAT, churn risk, etc.)
  • Create saved searches for operational workflows:
  • Low CSAT calls (e.g., CSAT < 3)
  • High churn risk calls
  • Escalations by reason
  • Train supervisors/QA on how to interpret explanations and drill down

Phase 6 — Expand

  • Add additional prebuilt tasks (topics, reason/outcome, sales insights)
  • Create tenant-specific custom insights (your business-specific fields)
  • Configure Auto QA if you have a QA program