Skip to main content

​
Post-Call Analysis

After every call, HubTalk AI runs an analysis pipeline that extracts structured data from the conversation. Configure analysis variables to capture the insights that matter to your business.

​
How It Works

  1. The call ends and a transcript is generated
  2. The LLM analyzes the conversation against your configured variables
  3. Results are stored and delivered via webhook and displayed in the dashboard
  4. Data can be exported as CSV from the Call Detail section

​
Analysis Variables

You define custom analysis variables in the agent’s Post-call Analysis settings. Each variable instructs the LLM to extract or compute a specific value from the conversation.

​
Variable Types

TypeDescriptionExample
TextFree-form text extractionCall summary, key topics discussed
SelectorPick from predefined options”interested” / “not interested” / “callback”
BooleanTrue/false determination”Did the caller agree to the offer?”
NumberNumeric extractionDebt amount mentioned, callback count

​
Built-in Analytics

Every call automatically includes:
FieldDescription
summaryLLM-generated summary of the conversation
sentiment_analysisCaller sentiment: positive, neutral, or negative
full_transcriptComplete text transcript with role labels
conversationStructured conversation split by role (agent/user)

​
Webhook Delivery

All analysis results are included in the webhook payload under analytics_data.reports.dynamic_reports. Each configured variable appears as a report entry with: 
{
  "name": "your_variable_name",
  "uuid": "report-uuid",
  "content": "LLM analysis output",
  "result": "extracted value"
}

​
Dashboard & Export

  • Analytics section — View aggregated metrics across calls
  • Call Detail section — Drill into individual calls with full transcripts and analysis
  • CSV Export — Download call details including all analysis variables