Skip to main content

​
Fine-tuning

Fine-tune your AI agent’s behavior by customizing the workflow prompt and configuring agent settings. This allows you to tailor the agent’s responses and actions to match your specific lending process and requirements.

​
Workflow Prompt

The workflow prompt is the main system prompt for the AI agent. You can:
  • Define agent behavior: Specify how the agent should interact with leads
  • Set priorities: Define what the agent should focus on first
  • Customize tone: Adjust the agent’s communication style
  • Add instructions: Provide specific guidance for different scenarios

​
Example Customizations

Prioritize document collection: 
Please focus on collecting all required documentation first before proceeding to other checklist items.
Custom follow-up instructions: 
For drop-off leads, wait 2-3 days before following up, and schedule follow-ups for 9-11 AM in the lead's local timezone.

​
Knowledge Base

Enhance the agent’s understanding by adding:
  • Company information: Your company background and values
  • Product details: Information about your loan products
  • Process documentation: Your lending process and requirements
  • FAQ content: Common questions and answers
This helps the agent understand your company better and provide more accurate, automated responses to leads.

​
Agent Behavior Settings

Configure how the agent behaves:
  • Response style: Professional, friendly, formal, etc.
  • Communication preferences: When to use email vs SMS
  • Escalation rules: When to escalate to human agents
  • Follow-up frequency: How often to follow up with leads

​
Testing Your Configuration

After making changes:
  1. Review the workflow prompt to ensure it matches your requirements
  2. Test with sample leads to verify behavior
  3. Monitor agent responses in the Timeline
  4. Adjust as needed based on results

​
Best Practices

  • Be specific: Clear, specific instructions work better than vague guidance
  • Set priorities: Clearly define what the agent should focus on
  • Update regularly: Keep the knowledge base current with latest information
  • Test changes: Always test configuration changes before deploying widely

Agent Settings

Configure agent communication settings

Conversation

Understand conversation handling

Autopilot

Enable automatic agent responses