# Best Practices

Use these practices to improve answer quality, reduce failed handoffs, and keep your chatbot useful over time.

## Start with accurate sources

Use current product documentation, policies, and support answers. Remove conflicting copies and verify the source is trained before evaluating its answers. Give each page a clear subject and keep important facts close to the question they answer.

## Define the agent’s boundaries

Describe who the agent helps, what it can answer, how it should respond when information is missing, and when to involve a teammate. Keep business facts in maintainable sources and behavior in Instructions.

## Test as a customer

1. Ask a common question in ordinary language.
2. Ask a follow-up that relies on the previous message.
3. Ask for a fact your sources do not contain.
4. Exercise a real action and check the result outside the chat.
5. Repeat the essential checks on mobile and the deployed channel.

## Review and improve

Use conversations, feedback, and knowledge gaps to choose the next correction. Recheck the same scenario after saving a change. Keep a short record of the expected answer and actual outcome so improvements can be measured.

- [Response quality](/docs/user-guides/quick-start/response-quality)
- [Knowledge gaps](/docs/user-guides/analytics-monitoring/knowledge-gaps)