Introduction

The "Conversations" component in wolkvox Studio allows the implementation of automated and dynamic service flows that understand and respond to customer intents in real-time, both through voice and chat.

Why Use the Conversations Component?

• Intelligent Automation: Reduces operational load by managing common requests without human intervention.
• Personalized Experience: Analyzes context and responds in a coherent and tailored manner.
• Flexibility: Compatible with different scenarios and sectors (support, sales, scheduling, etc.).

Use Cases

For example, this component can be used for scenarios such as:

• Automated Telephone Service: A customer calls a veterinary clinic to schedule an appointment or request information about pet care.
• Chatbot Support: A user interacts with a chatbot on WhatsApp to inquire about available products or services.
• AI-Based Voice Flows: A person communicates their request via voice, and the system automatically identifies their intent to direct them to the appropriate area.

Common Configuration (Voice and ChatBot)

Most of the options in the Conversations component are shared between Voice (IVR) and ChatBot routing points. These configurations define the general behavior of the artificial intelligence, the context of the conversation, and the logic for identifying intentions.

AI Model

The "AI Model" field allows you to select the type of model that will process customer interactions. This setting impacts the speed and accuracy of the analysis.

Available options:

• Fast: This model prioritizes response speed, which is ideal for flows where faster interaction is required, although it may sacrifice some accuracy. Use the Fast model when minimizing response times and computational load is a priority.
• Smart: This model prioritizes accuracy in intention analysis, which is useful for flows where more detailed and accurate responses are needed, although it may be slightly slower. Use the Smart model when a more precise analysis of customer intentions is required, especially in cases where responses must be highly personalized or complex.
• Reasoner: Uses a more advanced AI capable of processing complex reasoning and more elaborate language. Useful for sophisticated operations.
• Automatic: The system analyzes the customer's input and automatically chooses whether to use the Fast, Smart, or Reasoner model based on the context.

wvx Copilot Knowledge

This field allows linking the component to a domain trained in wvx Copilot, leveraging previously uploaded instructions and documents.

• If you select a domain, the AI will use that additional context to respond.
• If you choose "none", the system will not apply wvx Copilot knowledge.

Purpose, Conversation Tone

In this tab, you define:

• General purpose of the bot: Example: "Schedule medical appointments", "Provide technical support", "Resolve billing questions", or "Identify the reason for the customer's call to direct them to the correct area".
• Conversation tone: Professional, friendly, kind, formal, helpful, empathetic, direct, etc.

This purpose guides response generation and ensures consistency in the interaction.

Important Considerations

A good prompt structure is key for the AI to understand and respond correctly, directing the customer to the appropriate route and improving the experience.

Recommended Prompt Structure

• Categories and segments: Clearly define the categories and subtopics that the AI must handle.
• Keywords: Specify relevant terms for each intention.
• Intention codes: Use unique codes in brackets (e.g., [CODAPPT], [CODSALES]).
  • No spaces or special characters.
  • Mandatory for identifying intentions.
• Messages and responses: Write clear responses that include the intention code.
• Specific instructions: Indicate how to act in unforeseen cases (e.g., NO-MATCH).

Example of Prompt Structure

You are John Doe, the virtual assistant of wolkvox Pets. Your mission is to identify the customer's main need and collect key initial information to direct them correctly. Maintain a professional and empathetic tone.

Category: wolkvox Pets Veterinary Clinic

Segments:

1. APPOINTMENT SCHEDULING:
- Keywords: appointment, schedule, consultation, check-up.
- Action: If the customer shows intent to schedule an appointment, first ask for the pet's name, desired date, and time.
- Once you have all this information confirmed with the customer, say the following message and code: Give me a moment, I'll transfer you to the selected area [CODAPPT].

2. VACCINATION:
- Keywords: vaccines, vaccination, vaccine schedule.
- Action: If the customer mentions words related to vaccines, ask about the type of vaccine needed and confirm the information.
- Once confirmed, say the following message and code: Give me a moment, I'll transfer you to the selected area [CODVACCINES].

3. CARE AND FEEDING:
- Keywords: care, feeding, health, advice.
- Action: If the customer seeks information about care or feeding, ask what type of food they currently use for their pet and confirm the information.
- Once confirmed, say the following message and code: Give me a moment, I'll transfer you to the selected area [CODFEEDING].

4. EMERGENCIES:
- Keywords: emergency, urgency, immediate help.
- Action: If the customer mentions an emergency, request brief details about the situation and confirm the severity.
- Say the following message and code: I understand this is an emergency. I will connect you with a veterinarian immediately [CODEMERGENCY].

5. SPEAK TO AN AGENT:
- Keywords: speak to an agent, I want to talk to a human, I don't want to talk to a bot.
- Action: If the customer wants to speak to an agent, say the following message and code: Allow me a moment, I'll connect you with one of our advisors to assist you with your request [CODAGENT].

6. NO-MATCH (Unrecognized Response):
- Action: If you fail to understand the customer's response, request confirmation. Example: "I didn't understand your request. Could you try again? For example, you can write 'I want to schedule an appointment' or 'I want to speak to an agent'."

Additional context:
- Always use a friendly and professional tone.
- Prioritize emergency-related queries over other topics.
- Do not provide medical information without prior confirmation.
- Use available variables to tailor each response based on the customer's specific information. Address the customer by their first name, middle name, or last name occasionally so the customer feels that personalized attention. Example: "I’m happy to help, Daniel! Let's solve this together."

Intentions and Variable Extraction

This section allows you to define the intentions the AI must recognize and the variables that will store relevant information for the flow. It is essential for the Conversations component to direct the customer along the correct route and personalize the experience.

1. Go to the "Intentions and Variable Extraction" tab.
   • Key elements:
     • Intention (Intentions): Each intention represents a customer purpose (e.g., [CODAPPT], [CODSALES]).
     • words (Keywords): List of terms that help the AI identify the intention.
     • Variables: Fields to store extracted information (e.g., name, date, reason).
2. The "+" button allows you to add new intentions.
3. The table shows all the intentions defined in the "Conversations" component.

Clicking the "+" button opens a new configuration window where you can define the details of the intention:

1. "Intention" field:
   • Description: Here you specify the name of the intention. It must be written in brackets and without any special characters or spaces (e.g., [CODFEEDING]).
   • Example: [CODFEEDING]
   • Important: The intention name must be unique and clear to facilitate its identification within the flow.
2. "Comma-separated words for intentions" field:
   • In this field, list the keywords that the AI will use to detect the customer's intention. Words must be separated by commas.
   • Example: For the intention [CODFEEDING], you could write: care, feeding, health, advice.
   • Considerations:
     • Avoid duplicating words between different intentions to ensure the AI does not get confused.
     • Use specific and common keywords according to the type of customer.
3. "+" button: This allows you to add variables with their respective prompts so that the AI saves certain information in each variable created within the intention. Below this button is the variables table, which shows the variables added to the intention along with the prompts for each one.
4. Once you have the fields configured, click the "Add Intention" button.

After configuring the keywords, you can add variables to extract specific information from the conversation. This is done using the "+" button located below the variables table.

1. Variable name: Here you write the name of the variable that will be used to store the extracted information. Do not include the $ sign (the system will add it automatically). Variable names must remain in lowercase. Additionally, if you use any uppercase letters in the variable name, the system will automatically change them to lowercase.
   • Example: pet_name
2. Variable prompt: This field contains the prompt that instructs the AI on what information to extract from the conversation and save in the corresponding variable.
   • Example: Save the pet's name mentioned by the customer in this variable.
   • Note: The prompt must be specific and focused solely on data extraction. Do not include additional instructions such as responding to the customer, as this could cause execution errors.
3. Add variable: Clicking this button adds the variable with its respective prompt to the list of variables associated with the intention.

Complete Example of Intention Configuration

1. Intention: [CODFEEDING]
   • Intention name: [CODFEEDING]
   • Keywords: feeding, food, diet, nutrition, meal, ration, feeding recommendations, kibble, wet food.
   • Variables:
     • Variable 1:
       • Name: food_type
       • Variable prompt: Detect and save the type of food the customer mentions during the conversation, such as 'dry', 'wet', 'raw', etc.
     • Variable 2:
       • Name: preferred_brand
       • Variable prompt: Identify and save the food brand the customer prefers or mentions.
     • Variable 3:
       • Name: special_requirements
       • Variable prompt: Save any special requirements the customer mentions, such as allergies, specific dietary needs, or preferences.
2. Intention: [CODAPPT]
   • Intention name: [CODAPPT]
   • Keywords: appointment, schedule, consultation, check-up, reservation, schedule, availability, date, agenda, veterinarian.
   • Variables:
     • Variable 1:
       • Name: pet_name
       • Variable prompt: Save the name of the pet for which the appointment is being scheduled.
     • Variable 2:
       • Name: appointment_date
       • Variable prompt: Detect and save the date on which the customer wants to schedule the appointment.
     • Variable 3:
       • Name: appointment_time
       • Variable prompt: Identify and save the preferred time for the appointment.
     • Variable 4:
       • Name: appointment_reason
       • Variable prompt: Save the reason or purpose for which the customer is scheduling the appointment, such as 'annual check-up', 'vaccination', 'health issue', etc.

AI Response

In the "AI Response" field, you can obtain the variable that contains the entire response that the artificial intelligence provided to the customer.

Save Configuration

Don't forget to save your changes once you have fully configured the component.

Specific Configurations by Routing Point Type

Voice

When the Conversations component is used in Voice (IVR) routing points, additional configurations are enabled, allowing for a natural conversational experience through voice recognition (ASR) and speech synthesis (TTS).

TTS (Text-to-Speech)

TTS converts the system's responses into speech to interact with the customer. In this section, you can select the type of voice the system will use to "speak" to the customer.

Note:

• You have a wide variety of voice options available, including neural voices, which are more precise and natural, enhancing the customer's listening experience.
• The voice selected in this configuration will be used throughout the conversational flow, providing clear and professional responses.
• The appropriate voice selection can be adapted to the brand identity or the preferences of the target audience, ensuring a more personalized and pleasant interaction.

Interrupt TTS When Detecting Customer's Voice

This option allows the system to stop audio playback when it detects that the customer starts speaking, simulating a more natural conversation. However, this feature has important limitations:

• This checkbox allows you to configure whether the system should interrupt the text-to-speech (TTS) playback when it detects that the customer is speaking. Its activation or deactivation depends automatically on the flow structure.
  • Deactivation of interruption check: Occurs if the component has a No-MATCH route.
  • Activation of interruption check: Occurs if the component does not have a No-MATCH route.
• Recommendations:
  • If you want to ensure a structured and secure response, use a "No-MATCH" route and allow the TTS to play completely before listening to the customer.
  • Only consider using interruption if:
    • You want a smoother conversation experience.
    • You are sure the customer will interact in quiet environments.
    • The flow is short and voice detection must occur quickly.
  • Keep in mind: Interruption can be triggered by any sound, even background noise or nearby people. This can negatively affect the bot's understanding if executed in noisy environments.

ASR (Automatic Speech Recognition)

ASR converts the customer's voice into text so that the AI can analyze it. Here, you define the language the system will use for automatic speech recognition.

Note:

• It is important to select the language corresponding to your target audience to ensure accurate and smooth detection of customer responses.
• Correct ASR configuration ensures that the system understands customer requests, enabling effective and personalized interactions.
• The most popular languages include: Spanish, English, French, Portuguese, German, Italian, Hindi, Japanese, Korean, and Dutch.

ChatBot

Automatic Voice Note Recognition

For ChatBot routing points, the "Conversations" component includes the automatic detection and interpretation of voice notes sent by the end customer.

This functionality is activated proactively and does not require any additional manual configuration. When a user sends a voice message in the chat, the system automatically processes it to identify the customer's intention, just as it would with a text message.

Important Considerations

To ensure the Conversations component works correctly and provides a smooth experience for the customer, keep the following recommendations in mind:

1. Connect all intentions: Each created intention must be linked to an exit route in the flow.
   1. If the AI detects an intention (which has been created in the component configuration) without a defined route, the flow will reach a dead end and the interaction may end abruptly.
2. Create a NO-MATCH route: Always include a route to handle cases where the AI does not recognize the customer's intention.
3. Variable names in lowercase: Variable names must remain in lowercase. Additionally, if you use any uppercase letters in the variable name, the system will automatically change them so that all characters are in lowercase.