Skip to content

Conversation Analytics Lifecycle

This page explains the end-to-end flow from raw conversations to dashboards and searchable insights.

High-level flow

At a high level, Conversation Analytics follows this pipeline:

Conversation content
  ├─ Voice calls → Transcription → Transcript
  └─ Text channels (chat/email/tickets) → Normalized thread text
              ↓
         AI Tasks (enabled)
              ↓
      Custom Fields (stored results)
              ↓
Conversation Details · Dashboards · Search

Step 1: Collect conversation content

A conversation can include:

  • a voice call recording (audio)
  • a chat conversation
  • an email thread
  • a ticket with messages/notes

MiaRec associates the content with conversation metadata (agent, queue/team, timestamps, direction, etc.) so you can filter and segment analysis later.

Step 2: Convert to analyzable text

For voice calls, MiaRec uses transcription to produce a text transcript.

For text channels, MiaRec analyzes the conversation thread directly (when enabled).

If there is not enough text content, some insights may be skipped or returned as “unknown / insufficient evidence.”

Step 3: Run AI Tasks

An AI Task is a purpose-specific analysis definition that:

  • reads the transcript/thread text (and optionally metadata)
  • applies AI instructions (a prompt)
  • produces structured outputs

AI Tasks can be enabled/disabled and can be limited to certain conversations using filters (e.g., inbound calls longer than 15 seconds).

See: - AI Tasks and Prompts - Filters and Eligibility

Step 4: Store results in Custom Fields

MiaRec stores insight outputs in Custom Fields, such as:

  • CSAT (number)
  • Sentiment (category)
  • Top issue (multi-select)
  • Next action (text)
  • Reservation date (date)

Because results are stored in structured fields, they can be consistently used across the product:

  • dashboards and trend charts
  • drilldowns via clickable buckets
  • advanced search and filtering

See: - Custom Fields and Metrics

Step 5: Use insights in day-to-day workflows

Once stored, insights become usable signals for different teams:

  • CX leaders: trend CSAT and top issues
  • supervisors: drill into low scores and coach with evidence
  • QA teams: monitor Auto QA distributions
  • sales: track objections and competitor mentions

For examples, see: - Use Cases