Introduction

The NLP component of wolkvox Studio allows you to incorporate an Artificial Intelligence bot into your flow so it can answer customer questions or receive an order, process it, and return a result. Although the component is pre-trained with general knowledge, to make it respond in alignment with your operation, you need to train it with context and instructions (which you will see in the component as the Creative Area).

This component is part of the Cognitive group and is available in multiple routing types: Voice, Interaction, Chat, CRM + Webhook + cron + autoQAi + wvxCopilot, as well as Agent Scripting.

Configuration

1. Double-left-click on the component to open its configuration panel.
2. In the configuration panel, you will find:
   • “Example templates”: a dropdown selector to choose an existing template.
     • If you wish to create a new one, type the name and then use “Load template” to initialize it.
     • Recommendation: use templates per use case (for example: Support, Sales, Collections, Scheduling) to maintain consistency in responses and facilitate maintenance.
   • “Load template”: action to load the selected template.
     • Select an existing template in “Example templates” and click “Load template.”
   • In “Creative area” you must write the context the bot should consider and the instructions to generate the expected response. What to include here:
     • Bot role (what it is and what it does).
     • Objective (what it must resolve).
     • Response rules (tone, format, limits, what to do if it doesn't know).
     • Operation information (products, hours, policies, steps).
     • Flow variables (if you need it to use customer data, orders, cases, etc.).
     • Best practice: the more specific the context is (without making it unnecessarily long), the more consistent the bot's behavior will be.
   • Below the Creative Area, you will see the “JSON Format” checkbox. Use it when:
     • You need the result to be consumed by other components or integrations.
     • You want to extract specific fields (for example: intent, category, priority, summary, next_step, etc.).
     • Recommendation: if you enable JSON, also indicate it in the Creative Area; specify which keys you expect and provide an example of the format to maintain consistency.
   • Test the bot before saving (“Test” button).
     • Make sure the context is ready in the “Creative area.”
     • Click “Test.”
   • Review the output in “Result” to validate if it responds as expected.
     • Tip: test real-world scenarios (frequently asked questions + edge cases). If you notice ambiguity, adjust the instructions in the Creative Area.
   • Further down you will see “Response variable,” which already has a defined name (for example, $res_NLPG). This variable is where the component saves the result generated by the bot. Key points:
     • The variable name cannot be changed within this component.
     • Use that variable later in the flow to: Show the response to the customer (chat/voice), make decisions, or send the result to integrations or records.
   • Click on “Save” to apply the changes.