Voice Agent Field Extractors

Overview

Field Extractors for voice agents work identically to text agent field extractors - they automatically capture information from phone conversations and save it to contact fields. As your AI voice agent talks with callers, it listens for specific information and updates contact records automatically.

Key Benefit: No manual data entry after calls. Information discussed is automatically saved.

What Voice Field Extractors Do

Example Scenario:

Agent: "What service do you need help with?" Caller: "I need a new HVAC system installed at 123 Main Street in Austin, Texas, 78701"

With Field Extractors configured:

• service_description → "new HVAC system installed"
• street_address → "123 Main Street"
• city → "Austin"
• state → "Texas"
• zip_code → "78701"

All saved automatically during the call.

When to Use Field Extractors

Definitely Use When:

• Voice agent asks questions to gather information
• Information is critical for follow-up (addresses, service details)
• Data needed for booking or service delivery
• Want to avoid post-call data entry

Examples:

• ✅ Agent asks for address → Extract street_address, city, state, zip_code
• ✅ Agent asks about service needed → Extract service_description
• ✅ Agent gathers project details → Extract to custom fields
• ✅ Caller mentions email → Extract email
• ✅ Caller provides business name → Extract business_name

Configuration

Voice agent field extractors are configured identically to text agent field extractors:

Field Selection:

• Choose from standard contact fields
• Or select custom fields

What to Extract:

• Describe what information to capture
• Be specific and include examples
• Example: "the caller's street address. Ex: '123 Main St'"

Allow Overwrite:

• Enabled: New data overwrites existing values
• Disabled: Only save if field is empty
• Default: Enabled (recommended)

See: Field Extractors (Text Agents) - Same configuration applies to voice

Common Voice Field Extractor Setups

Setup 1: Service Call Information

Service Description:

• Field: service_description
• Description: "the caller's description of the service they need or the issue they're experiencing"

Setup 2: Complete Address

Street Address:

• Field: street_address
• Description: "the street address where service is needed. Ex: '123 Main St'"

City:

• Field: city
• Description: "the city where service is needed. Ex: 'Austin' "

State:

• Field: state
• Description: "the user's state. Ex: 'TX' (Alway use two letter postal abbreviation)"

Zip Code:

• Field: zip_code
• Description: "the 5-digit zip code. Ex: '78701'"

Setup 3: Business Information

Business Name:

• Field: business_name
• Description: "the name of the caller's company or business"

Email:

• Field: email
• Description: "the caller's email address with no whitespace. Ex: '[email protected]' "

Voice-Specific Considerations

Verbal vs Written

Challenge: Voice data may be less structured than text

Solutions:

• AI is good at understanding verbal addresses, numbers, etc.
• Include format examples in descriptions
• Review extracted data periodically
• Be specific with AI about normalizing formats (e.g., "seventy-eight thousand seven oh one" → "78701")

Caller ID vs Provided Number

Caller ID:

• Automatically captured as contact phone
• May be different from preferred callback number

Extract separate "preferred phone" if:

• Caller mentions different callback number
• Calling from office but wants cell callback
• Create custom field for "preferred_contact_number" (Settings > Custom Fields)

Confirmation is Good

In voice agent instructions, confirm extracted info:

• "Just to confirm, that's 123 Main Street in Austin, 78701?"
• Improves accuracy
• Builds caller confidence

Best Practices

Match to Agent Instructions

If agent asks for specific information, extract it:

• Agent asks for address → Extract address fields
• Agent asks about service → Extract service_description
• Agent gathers timeline → Extract custom field

Use for Call Documentation

Extract key details for post-call reference:

• What was discussed
• Commitments made
• Information provided
• Special requests

Don't Over-Extract

Only extract what you need:

• ✅ Critical service information
• ✅ Contact details
• ✅ Booking requirements
• ❌ Every possible field "just in case"

Review Accuracy Periodically

Voice extraction can have quirks:

• Check contact records after calls
• Verify addresses are correct
• Confirm phone numbers formatted properly
• Adjust descriptions if needed

Monitoring Field Extractors

Voice Agent Logs

Access: AI Agents page > Voice Logs tab

What You See:

• Which field extractors fired during calls
• What values were extracted
• Whether fields were updated
• Call transcripts showing where information was mentioned

Use For:

• Verifying extraction accuracy
• Debugging missed extractions
• Optimizing descriptions
• Quality assurance

Troubleshooting

Problem: Field extractor not firing during calls

Solution:

• Ensure extractor is enabled
• Verify caller actually provided that information
• Make description more specific or flexible
• Review call transcripts in voice logs
• Test with voice agent

Problem: Wrong data being extracted

Solution:

• Refine "What to Extract" description
• Add format examples
• Be more specific about what to capture
• Review call logs for patterns

Problem: Address extracted incorrectly

Solution:

• Add address format example to description
• Agent should confirm address verbally
• Consider having agent ask for components separately (street, city, zip)

Problem: Phonenumbers/emails garbled

Solution:

• Agent should repeat back for confirmation
• Add format examples to extraction description

Related Features

• AI Voice Agents Overview - Main voice agent guide
• Voice Conditional Tools - Take actions during calls
• Field Extractors - Same concept for text agents
• Contacts - View extracted data in contact records