This article guides you through the process of validating that an advanced AI agent is ready for launch. Resource links for each item point you to information that can help you check or troubleshoot. Some resources are specific to AI agents based on expressions, zero-training, or agentic AI.
Checklist
Item | Description | Resources |
---|---|---|
The tone of voice is correct and consistent. | All replies from the AI agent are aligned with its persona. | |
Knowledge sources are imported successfully. | The AI agent has access to the information it needs. | |
All use cases or intents are in place. | You added all the use cases or intents you needed. Fill gaps with use case suggestions or the intent assistant. | |
The correct actions are set up. | All desired AI agent-level, use case-level, and block-level actions have been configured. | |
The correct CRM actions are being triggered. | Actions are triggered as expected in your integrated CRM, like Zendesk Support or Sunshine Conversations. | |
All replies are working as intended. | Behaviors have been validated. For example, blocks are triggered correctly, and conversation flows make sense and work from start to finish, with no lingering placeholder text or TBDs. | |
The correct use cases/intents and replies are triggered. | Conversation logs show the expected use cases or intent and replies. If not, use the logs for troubleshooting. | |
Conditional blocks are working correctly. | Review dialogues and see if all conditional blocks are working as expected. | |
Entities are recognized. | All entities are recognized in conditional blocks, in all replies, for all languages you’re using. |