Introduction

The creation of effective conversational agents in platforms like wolkvox Studio largely depends on how intents and variables are designed and managed. These are the fundamental pieces that allow AI systems to understand, process, and respond to customer needs accurately and in a personalized manner.

Importance of Intents and Variables

• Intents: Intents allow identifying the purpose or need expressed by the customer within the conversation. For example: requesting an appointment, asking for technical support, or making a purchase.
• Variables: Variables are used to capture specific information within each intent, such as the date of an appointment, the customer's name, the type of problem reported, or any other relevant data.

Architecture of Intents

Intents are the core of conversational agents in wolkvox Studio, as they represent the different actions or purposes that the customer can express during the interaction. A good definition of intents allows the AI to correctly interpret what the customer wants to achieve.

Definition and Purpose

• Each intent corresponds to a category of request (e.g., scheduling an appointment, requesting information, filing a complaint).
• They allow guiding the conversation towards the correct flow path in the routing point design.

Recommended Nomenclature

To facilitate management, it is recommended:

• Use unique and descriptive names.
• Use prefixes or codes that identify the purpose:
  • [CODCITA] → appointment scheduling.
  • [CODVENTA] → sales management.
  • [CODSOPORTE] → technical support.
• Avoid spaces, special characters, and ambiguous names.

Keywords and Synonyms

Each intent should be configured with a list of words or phrases that the customer might use.

• Example: [CODCITA] can include "appointment, schedule, book, reserve, consultation."
• It is important to add synonyms and natural language variations to increase recognition accuracy.
• It is recommended to maintain a balanced list: neither too short (may fail in detection) nor excessively long (may cause confusion with other intents).

Variable Management

Variables in wolkvox Studio allow capturing and storing relevant information provided by the customer during the conversation. These variables are essential for personalizing the experience, validating data, and maintaining flow continuity.

Creating Variables in the Interface

In the "Intents and Variable Extraction" tab within the Conversations component, it is possible to:

• Define the variable name.
• Establish an extraction prompt, i.e., a clear instruction so that the AI knows what data to capture.
• Save and associate the variable with a specific intent.

Variable Extraction Prompts

The prompt should be clear, brief, and specific.

• Correct example:
  • Variable: appointment_date
  • Prompt: "Detect and save the date requested by the customer."
• Incorrect example:
  • Ambiguous prompt like "date" or "time," as it does not adequately guide the AI.

Recommendation: Always use direct language in the prompt to improve extraction accuracy.

Relationship Between Intents and Variables

Intents and variables work together in wolkvox Studio's conversational agents.

• The intent identifies what the customer wants to do.
• The variables extract and store the necessary information to fulfill that intent.

How to Associate Variables with Intents

In the Conversations component configuration:

• The intent is created, and keywords are defined.
• One or more variables can be linked to that intent.
• Each variable will have a specific prompt that will guide data extraction.

This ensures that every time the intent is detected, the system automatically captures the relevant information.

Practical Example: Appointment Scheduling

• Intent: [CODCITAS]
  • Keywords: appointment, schedule, book, consultation, review.
  • Associated variables:
    • appointment_date → "Detect and save the date requested by the customer."
    • appointment_time → "Detect the time mentioned in the request."
    • appointment_reason → "Save the reason for the appointment."

In this example, the intent [CODCITAS] activates the scheduling flow, and thanks to the variables, the system obtains the necessary data to complete the scheduling.

Validations and Intent Routing

Defining intents and variables is not enough: it is necessary to validate the captured data and correctly route the conversation according to the detected purpose.

Data Validations

Validations ensure that the extracted information is consistent and useful.

Examples of recommended validations:

• Dates: Check that the captured date corresponds to a valid format or is within an allowed range.
• Emails: Verify that they contain the "@" character and a valid domain.
• Phones: Validate that they have the correct length (e.g., 10 digits in Colombia).
• ID Numbers: Confirm that they meet the expected format in each country.

If data does not pass validation, the flow should ask the customer to repeat it.

Using Intent Codes for Routing

Each intent should be associated with a unique code indicating which route the conversation should take.
Examples:

• [CODCITAS] → Redirects to the scheduling flow.
• [CODSOPORTE] → Opens the technical diagnosis route.
• [CODVENTAS] → Activates the sales route.
• [CODAGENTE] → Transfers the interaction to a human agent.

This ensures that the system responds in a structured and coherent manner according to the detected context.

Practical Examples

The following cases show how to apply intents and variables in common scenarios for conversational agents in wolkvox Studio.

Case: Technical Support on WhatsApp

• Intent: [CODSOPORTE]
  • Keywords: damage, failure, error, not working, problem.
  • Associated variables:
    • problem_type → "Detect the type of problem mentioned by the customer."
    • equipment_model → "Save the model or reference of the affected equipment."
    • error_description → "Extract the detailed description of the error."
• Route: If the problem_type variable contains "internet," route to the connectivity support flow; if it contains "bill," route to the billing clarification flow.

Case: Scheduling Medical Appointments

• Intent: [CODCITAS]
  • Keywords: appointment, consultation, book, schedule, review.
  • Associated variables:
    • patient_name → "Save the patient's name."
    • appointment_date → "Extract the requested appointment date."
    • appointment_time → "Capture the appointment time."
    • appointment_reason → "Detect the reason for the medical appointment."
• Route:
  • If the appointment_date is not available in the schedule, offer nearby alternatives.
  • If the appointment_reason contains the word "urgent," transfer directly to a human agent.

Best Practices

The correct use of intents and variables in wolkvox Studio ensures that conversational agents are clearer, more efficient, and more natural. Below are some key recommendations:

• Design clear and non-redundant intents: Avoid creating intents with overly similar keywords, as they can cause confusion in detection.
• Cover natural language variations: Include synonyms, abbreviations, and common phrases that customers might use in each intent.
• Use specific extraction prompts: A clear prompt improves AI accuracy. Example: "Detect the appointment date" is more useful than just "date."
• Always validate captured data: Before using them in the flow, confirm that they meet the expected format (date, email, phone number).
• Maintain consistency in variable names: Use lowercase and descriptive names, without special characters. Example: appointment_date, customer_name.
• Define a NO-MATCH route: Always create a route to handle cases where no intent matches, preventing the customer from being left without a response.
• Associate each intent with a unique code: This facilitates routing and traceability in reports.

Quality Checklist

Before publishing a flow with intents and variables in wolkvox Studio, review the following points to ensure their correct functioning:

[ ] Unique and descriptive name.

[ ] Keywords include enough synonyms and natural language variations.

[ ] Variables are correctly defined and have clear, specific prompts.

[ ] Validations are configured for sensitive data such as dates, emails, and phones.

[ ] Each intent is associated with a unique routing code.

[ ] There is a NO-MATCH backup route to handle non-matching cases.

[ ] Variables follow a consistent naming convention (e.g., lowercase, no special characters).

[ ] Practical examples were tested with real customer phrases.

[ ] The entire flow avoids redundancies and maintains a natural conversation.