Training Your Chatbot
Fine-Tune Your Chatbot's Responses
Why Fine-Tune?
Even with good training data, your chatbot may:
- Miss nuances in certain questions
- Provide outdated information
- Use the wrong tone or format
- Lack context for specific scenarios
Fine-tuning addresses these issues by letting you teach the chatbot exactly how to respond.
Using Updated Answers
1
Go to Training & Content
Click Training & Content in the sidebar.
Click on Updated Answers.
Click Add Updated Answer.
Type the question or phrase that triggered the wrong response.
5
Provide the correct answer
Write the response you want your chatbot to give.
Click Save to apply the correction.
Best Practices
Review Conversation Logs
Regularly check your chatbot’s conversations in the Inbox to identify:
- Frequently asked questions with poor answers
- Topics where the chatbot seems confused
- Missed opportunities to provide helpful information
Be Specific
When adding updated answers:
- Match the exact phrasing visitors use
- Include common variations of the question
- Provide complete, helpful responses
Test Your Changes
After fine-tuning:
- Use Test Agent to verify the improvement
- Monitor real conversations to confirm it’s working
- Iterate if needed
How It Works
For broader changes, update your AI instructions in Settings:
- Tone: Make responses more formal or casual
- Length: Set preferred response length
- Behavior: Define how to handle certain topics
- Restrictions: Limit what the chatbot will discuss