Skip to main content

Insight Templates

Insight templates define what information to extract from call transcripts. Each template specifies the data to capture and how it should be returned.

Template Components​

Every insight template has:

  • Name: Clear, descriptive identifier (e.g., "Lead Quality Score")
  • Instructions: Detailed rules for what to extract and how
  • Output Type: Text, Number, Boolean, Object, or Array
  • Category: Sales, Support, Summary, or Custom
  • Description: Explains what the insight captures (optional but recommended)

Pre-Built Templates​

Voka AI provides ready-to-use templates across multiple categories:

Required Template​

Call Summary (Text)

  • Automatically included with every assistant
  • Cannot be removed
  • Generates concise conversation overview
  • Powers call log preview text
  • Displays with purple highlight and "Required" badge

Sales Templates​

Lead Quality Score (Number: 0-100)

  • Measures buying intent and qualification level
  • Higher scores indicate stronger prospects
  • Color-coded in call logs: Green (80+), Purple (50-79), Gray (<50)

Purchase Intent (Text)

  • Returns: "High", "Medium", "Low", or "None"
  • Identifies readiness to buy

Budget (Number or Text)

  • Captures mentioned budget amounts
  • Can extract ranges or specific numbers

Decision Timeline (Text)

  • When customer plans to make purchase
  • Examples: "Next month", "Q1 2026", "Immediate"

Support Templates​

Customer Sentiment (Text)

  • Returns: "Positive", "Neutral", or "Negative"
  • Helps identify satisfaction levels

Issue Category (Text)

  • Classifies the support request type
  • Examples: "Billing", "Technical", "General inquiry"

Resolution Status (Text)

  • Whether issue was resolved on the call
  • Returns: "Resolved", "Pending", or "Escalated"

Summary Templates​

Action Items (Array)

  • List of follow-up tasks from the conversation
  • Example: ["Send proposal", "Schedule follow-up", "Call references"]

Key Details (Object/JSON)

  • Structured data with multiple fields
  • Can include: customer name, budget, timeline, pain points
  • Example: {"name": "John Smith", "budget": 5000, "timeline": "next month"}

Creating Custom Templates​

If pre-built templates don't fit your needs, create custom insights:

Step-by-Step​

  1. Go to Insights in main navigation
  2. Click Create Custom Insight
  3. Fill in the form:
    • Name: Clear identifier (e.g., "Property Size")
    • Description: What this insight captures
    • Instructions: Detailed extraction rules (see below)
    • Output Type: Text, Number, Boolean, Object, or Array
    • Category: Sales, Support, Summary, or Custom
  4. Click Create
  5. Add to any assistant via their Insights tab

Writing Effective Instructions​

Use this template for clear instructions:

Extract [specific information] from the conversation.

Look for:
- [Example 1]
- [Example 2]
- [Example 3]

Return: [exact format]
If [edge case], return [default value]

Custom Template Examples​

Example 1: Property Size

Name: Property Size
Instructions: Extract the property size mentioned by the customer.
Look for:
- Square footage ("2,500 sq ft", "3000 square feet")
- Lot size ("half acre", "5000 sq ft lot")
- Number of rooms ("3 bedroom", "2 bath")
Return: Numeric value only (e.g., 2500)
If not mentioned, return 0
Output Type: Number

Example 2: Competitor Mention

Name: Competitor Mentioned
Instructions: Identify if the customer mentioned any competitors.
Look for:
- Company names of direct competitors
- Phrases like "I'm also considering...", "compared to..."
Return: Company name if mentioned, or "None" if no competitor discussed
Output Type: Text

Example 3: Appointment Preferences

Name: Appointment Preferences
Instructions: Extract customer's scheduling preferences.
Look for:
- Preferred days of week
- Preferred time of day
- Any constraints or requirements
Return: JSON object with: {"day": "...", "time": "...", "notes": "..."}
If not mentioned, return empty object {}
Output Type: Object

Adding Custom Templates to Assistants​

  1. Go to AI Assistants → Select assistant → Insights tab
  2. Click Add More Insights
  3. Find your custom template in the list
  4. Select and add to assistant

Output Types​

Text​

For categories, names, descriptions

Example: "High", "Customer support issue"

Number​

For scores, quantities, amounts

Example: 8, 5000

Boolean​

For yes/no, true/false

Example: true, false

JSON​

For structured data

{
  "name": "John Doe",
  "budget": 5000,
  "timeline": "Q1 2025"
}

Array​

For lists

["Send proposal", "Schedule follow-up", "Call references"]

Best Practices​

Do:

  • Start with 2-5 essential insights (Call Summary is automatic)
  • Use pre-built templates when available
  • Test with 10-20 calls before going live
  • Write specific, detailed instructions for custom insights
  • Handle edge cases (e.g., "not mentioned", "unclear")
  • Specify exact output format expectations
  • Include example values in instructions

Don't:

  • Add too many insights at once (slows processing)
  • Use vague or ambiguous instructions
  • Skip testing custom insights thoroughly
  • Forget to specify what to return when data isn't present
  • Create overly complex nested objects unless necessary

Tips for Accurate Extraction​

  1. Be Specific: Instead of "Extract budget", use "Extract the dollar amount the customer mentioned as their budget. Return numeric value only."

  2. Provide Examples: Show what you're looking for with quotes like "I have $5,000 to spend" or "around ten thousand dollars"

  3. Handle Missing Data: Always specify what to return when information isn't mentioned (0, "None", null, empty object)

  4. Test Variations: Test with different phrasings customers might use

  5. Review and Refine: Check extracted data regularly and update instructions to improve accuracy