Skip to main content
The Playground is a testing environment where you can interact with your AI agent, evaluate responses, and fine-tune settings before going live.

​
Accessing the Playground

  1. Go to Agents > select your agent
  2. Click Playground in the sidebar

​
Testing Conversations

Type messages in the chat interface to test your agent:

​
What to Test

  • Accuracy - Are answers factually correct based on training data?
  • Relevance - Does the agent understand the question?
  • Completeness - Are responses comprehensive but concise?
  • Tone - Does it match your configured tone settings?
  • Edge cases - How does it handle unclear or off-topic questions?

​
Evaluating Responses

After each response, consider:
  1. Is the information accurate?
  2. Would a customer find this helpful?
  3. Does it reflect your brand voice?
  4. Are there gaps in knowledge?

​
Improving Responses

If responses need improvement:
1

Identify the issue

Is it a knowledge gap, tone issue, or misunderstanding?
2

Update training data

Add Q&A pairs or documents to address gaps
3

Adjust settings

Modify tone, language, or instructions
4

Retrain

Click Retrain to update the agent
5

Test again

Verify the improvement in Playground

​
Session Management

Each Playground session is independent. To start fresh:
  • Click New Session to clear conversation history
  • The agent won’t remember previous test conversations

​
Tips

  • Test common customer questions first
  • Try variations of the same question
  • Test in the language your customers use
  • Have team members test independently
  • Document issues and improvements needed