Configuring Enjo
AI Actions
5min
AI Actions are a way to train Enjo AI Agents to interact with external applications via RESTful API calls or using custom code. They convert natural language text into data retrieval (e.g. query an HRMS application) or take action in another application (e.g. reset password in Okta).
- AI Agents decide which AI Action is relevant to the user’s question
- AI Action typically takes confirmation from the user to execute the action
- AI Action queries the user for any missing information to execute the action
- AI action executes the action and confirms to the user in natural language
There are 2 types of AI Actions - API Action and Code Action. API Actions use an external API directly to execute the action in another application. Code Action uses NodeJS code to execute the action in the external application.
Adding relevance threshold for the AI Action :
A score measures the confidence level, and teams can customize the threshold to meet their specific requirements.
Key Benefits:
- Triggers AI actions based on query match percentage: Ensures actions are only activated when a strong match improves response relevance.
- Reduces accidental activations: Lowers the chance of inappropriate or mistaken AI responses.
- Improves AI accuracy in customer support: Boosts the precision and reliability of AI-driven responses, leading to better customer interactions.
- AI Agent is not calling the AI action or calling the wrong AI Action - Make sure the descriptions of the AI Actions are clear. Include as many variations of user questions in “Question variations” including difficult/ambiguous versions.
- AI Action is not using the parameters correctly - Add or modify the descriptions of the parameters in the AI Action
- AI Action is not working or giving an error - use Postman or a similar 3rd party service to debug the API
Updated 03 Dec 2024
Did this page help you?