Custom Conversation Extraction

​

What is Custom Conversation Extraction?

Custom Conversation Extraction allows you to define exactly what information you want Droxy to extract and analyze from every conversation. Instead of relying only on default analysis, you can create a custom schema that automatically extracts business-critical data points tailored to your specific needs. This powerful feature uses AI to intelligently identify and extract the information you specify, turning every conversation into structured, actionable data that you can filter, analyze, and act upon.

​

How to Configure Custom Conversation Extraction

1. From your agent’s dashboard, click the Advanced section
2. Scroll to the Custom Conversation Extraction accordion
3. Click Add Field to create a new extraction field
4. Configure your custom analysis fields (see examples below)
5. Click Save to apply your configuration

Once configured, Droxy will automatically analyze all future conversations with your agent and extract the specified information.

​

Field Types and Examples

You can create extraction fields with different data types to capture various kinds of information:

​

String Fields

Extract text-based information with predefined values or free-form text. Example: Lead Quality

• Field Name: Lead Quality
• Description: “Quality of the lead for sales purposes based on budget, timeline, and fit indicators”
• Type: String with predefined values
• Values: hot, warm, cold, not_applicable

Example: Product Mention

• Field Name: Product Mention
• Description: “The specific product, service, or feature name discussed in the conversation”
• Type: String (free-form)

​

Enumerated Values

Create fields with specific options to categorize conversations consistently. Example: Urgency Level

• Field Name: Urgency
• Description: “How urgently this conversation needs attention or follow-up”
• Type: Enum
• Values: low, medium, high, critical

Example: Department Routing

• Field Name: Department
• Description: “Which department should handle this conversation”
• Type: Enum
• Values: sales, support, billing, technical, general

​

Number Fields

Extract numerical data for quantitative analysis and scoring. Example: Satisfaction Score

• Field Name: Satisfaction Score
• Description: “Satisfaction score on a scale of 1 to 10 based on the customer’s tone and sentiment”
• Type: Number (1-10)

Example: Budget Mentioned

• Field Name: Budget Amount
• Description: “Budget or price range mentioned by the customer”
• Type: Number

​

Boolean Fields

Extract yes/no information for binary classification. Example: Demo Requested

• Field Name: Demo Requested
• Description: “Whether the customer requested a product demonstration”
• Type: Boolean

Example: Competitor Mentioned

• Field Name: Competitor Mentioned
• Description: “Whether the customer mentioned using or considering a competitor”
• Type: Boolean

​

Array Fields

Extract multiple values or items from a single conversation. Example: Objections Raised

• Field Name: Objections Raised
• Description: “Specific concerns, hesitations, or objections expressed by the customer”
• Type: Array of strings

Example: Pain Points

• Field Name: Pain Points
• Description: “Specific problems or challenges mentioned by the customer”
• Type: Array of strings

​

What You Can Unlock

Custom Conversation Extraction opens up powerful possibilities for your business:

​

Sales Intelligence

• Qualify Leads Automatically: Extract lead quality, budget, timeline, and decision-making authority to prioritize follow-ups
• Track Product Interest: Identify which products or features customers are most interested in
• Detect Buying Signals: Capture urgency indicators, budget mentions, and timeline references
• Competitive Analysis: Track competitor mentions and understand why customers are comparing options

​

Customer Success

• Measure Satisfaction: Automatically score customer satisfaction based on conversation tone and sentiment
• Identify At-Risk Customers: Flag conversations with negative sentiment or complaint indicators
• Track Feature Requests: Capture and categorize product feedback and feature requests
• Monitor Support Quality: Extract resolution status, response quality, and customer satisfaction

​

Marketing Insights

• Understand Customer Needs: Extract pain points, challenges, and goals mentioned in conversations
• Track Campaign Performance: Identify how customers heard about you and which campaigns drive conversations
• Content Opportunities: Discover frequently asked questions that could become blog posts or FAQs
• Persona Development: Build detailed customer personas based on extracted conversation data

​

Operations and Analytics

• Filter and Segment: Use extracted data to filter conversations in your inbox by any custom field
• Export for Analysis: Export conversations with custom fields to analyze trends and patterns
• Automate Workflows: Use extracted data to trigger automations (via Zapier or API integrations)
• Report to Stakeholders: Create dashboards and reports based on structured conversation data

​

Best Practices and Tips

Start Simple: Begin with 3-5 essential fields that align with your most important business metrics. You can always add more later. Use Clear Descriptions: The AI uses your field descriptions to understand what to extract, so be specific and detailed about what you’re looking for. Predefined Values Work Best: For categorical data, provide predefined values rather than free-form text. This ensures consistency and makes filtering easier. Review and Iterate: Check your extracted data regularly to ensure the AI is capturing information accurately. Refine your descriptions if needed. Export for Deeper Analysis: Regularly export your conversation data with custom fields to spreadsheets for trend analysis and business intelligence. Consider Your Team’s Needs: Involve sales, support, and other teams when designing extraction fields to ensure you’re capturing data that everyone can use.