autoch.at Documentation

Appearance

Conversation Intelligence ​

Conversation Intelligence automatically analyzes your customer conversations to help you understand what's working, what's not, and where you can improve.

What is Conversation Intelligence? ​

Conversation Intelligence is an AI-powered analytics system that:

  • Analyzes conversations to extract insights, objections, and patterns
  • Tracks response times and their impact on outcomes
  • Identifies missed follow-ups that need attention
  • Surfaces common objections and how well they're handled
  • Provides actionable insights to improve your AI assistant

All analysis happens automatically in the background—no manual work required.

Key Features ​

Response Time Analysis ​

See how quickly your AI responds and how that affects customer satisfaction:

  • Response time buckets: 0-5s, 5-15s, 15-30s, 30s+
  • Qualification rates by response time
  • Channel breakdown: Compare response times across email, SMS, chat, voice

Use this to:

  • Identify channels where response times are too slow
  • Understand the relationship between speed and outcomes
  • Set response time goals

Missed Follow-ups Queue ​

Get a prioritized list of conversations that need attention:

  • Stalled conversations: No activity for 48+ hours
  • Unanswered questions: Customer asked a question but got no response
  • Incomplete information: Missing details needed to proceed
  • Promised callbacks: Follow-up commitments that haven't happened

Each follow-up shows:

  • Reason: Why it needs attention
  • Priority: Low, medium, or high
  • Last activity: When the conversation last had activity
  • Channel: Where the conversation is happening

Objection Analysis ​

See what objections customers raise and how well your AI handles them:

  • Common objections: Price, timing, competition, trust, need, authority
  • Frequency: How often each objection appears
  • Addressed rate: How often objections are successfully handled
  • Example conversations: See real examples (with PII removed)

Use this to:

  • Identify objection patterns
  • Improve objection handling in your playbooks
  • Train your team on common objections

Intent & Risk Classification ​

Conversations are automatically classified by:

  • Intent: Question, complaint, request, feedback, support, sales, other
  • Risk level: Low, medium, high, critical
  • Urgency: 1-5 scale
  • Confidence: How confident the classification is (0.00-1.00)

These classifications are stored as first-class fields, making it easy to:

  • Filter conversations by risk level
  • Prioritize high-risk conversations
  • Route conversations based on intent
  • Build reports and dashboards

How It Works ​

Automatic Analysis ​

Conversation Intelligence runs automatically:

  • Daily analysis: Analyzes conversations from the past 24 hours
  • Background processing: Runs without impacting your system performance
  • Incremental updates: Only analyzes new conversations

Analysis Process ​

  1. Data Collection: Gathers conversations from all channels
  2. Response Time Calculation: Measures time between customer messages and AI responses
  3. Follow-up Detection: Identifies conversations that need attention
  4. Objection Extraction: Uses AI to identify and categorize objections
  5. Intent Classification: Classifies conversations by intent and risk
  6. Insight Generation: Creates actionable insights and recommendations

Version Tracking ​

Every analysis includes version metadata so you can track:

  • Model: Which AI model was used (e.g., "gpt-4o-mini")
  • Provider: Which provider (e.g., "openai")
  • Prompt Version: The analysis prompt version (e.g., "ci.objections.v1.0")
  • Code Version: The code version when analysis ran (e.g., "git:8f3c9c2")
  • Generated At: When the analysis was performed

This helps you:

  • Compare results across different analysis versions
  • Debug issues by knowing exactly what code/model was used
  • Track improvements over time

Using Conversation Intelligence ​

Accessing the Dashboard ​

  1. Go to Conversation Intelligence in the navigation
  2. You'll see:
    • Summary: High-level metrics and recent insights
    • Response Times: Charts showing response time vs. outcomes
    • Follow-ups: Queue of conversations needing attention
    • Objections: List of common objections and handling rates

Filtering Results ​

Filter by:

  • Date range: Last 7 days, 30 days, 90 days, or custom range
  • Channel: Email, SMS, chat, voice
  • Playbook: See results for specific playbooks (if playbook attribution is available)

Resolving Follow-ups ​

When you resolve a follow-up:

  1. Click on the follow-up in the queue
  2. Review the conversation
  3. Click Resolve
  4. Optionally add a resolution note
  5. The follow-up is marked as resolved and removed from the queue

Viewing Objection Details ​

Click on an objection to see:

  • Frequency: How many times it appeared
  • Addressed rate: Percentage successfully handled
  • Example conversations: Real examples (PII removed)
  • Category: Type of objection (price, timing, etc.)

Understanding the Metrics ​

Response Time Buckets ​

  • 0-5s: Very fast responses (ideal for chat and SMS)
  • 5-15s: Fast responses (good for most channels)
  • 15-30s: Moderate responses (acceptable for email)
  • 30s+: Slow responses (may need optimization)

Qualification Rates ​

The percentage of conversations that resulted in qualified leads, broken down by response time bucket. Use this to see if faster responses lead to better outcomes.

Objection Addressed Rate ​

The percentage of times an objection was successfully handled (customer continued the conversation or took the desired action). Higher is better.

Risk Levels ​

  • Low: Normal conversation, no special handling needed
  • Medium: Some concern, monitor closely
  • High: Significant concern, may need escalation
  • Critical: Urgent issue, immediate attention required

Best Practices ​

Regular Review ​

  • Check the follow-ups queue daily
  • Review objection trends weekly
  • Analyze response time patterns monthly

Action on Insights ​

  • High objection frequency: Update your playbook to better handle that objection
  • Slow response times: Optimize your AI configuration or add more resources
  • Many missed follow-ups: Review your follow-up processes

Using Risk Levels ​

  • Set up alerts for high-risk conversations
  • Route critical conversations to human agents immediately
  • Review medium-risk conversations regularly

Troubleshooting ​

No Data Showing ​

  • Wait for analysis: Analysis runs daily, so new conversations may not appear immediately
  • Check date range: Make sure your date range includes analyzed conversations
  • Verify conversations exist: Ensure you have conversations in the selected time period

Follow-ups Not Appearing ​

  • Follow-ups only appear for conversations that meet specific criteria (e.g., stalled for 48+ hours)
  • Check that conversations are in the correct status
  • Verify the conversation hasn't already been resolved

Objections Seem Incorrect ​

  • Objection extraction uses AI and may occasionally misclassify
  • Review example conversations to understand the classification
  • Common objections are aggregated, so minor variations are grouped together

autoch.at Documentation

Copyright © 2024 autoch.at