Introduction

Agentic Studio is a wolkvox Studio component that allows you to design and modify flows using natural-language instructions. Instead of manually building all the logic using different components on the canvas, the administrator can explain through text or voice what they need the flow to do, and Agentic Studio interprets the requirement and internally builds the necessary logic.

The component understands the capabilities available within the type of Routing Point where it is being used and can integrate the logic and functionality of other components to fulfill the request. In this way, it is possible to go from a functional need expressed in natural language to a flow structure that is ready to be reviewed, adjusted, and tested. This capability was originally designed for IVR creation, making it possible to generate logic, validations, transfers, and more complex processes without having to develop them from scratch.

Currently, Agentic Studio can be used in both Voice Flow and Chat Flow Routing Points and is located within the Cognitive component group in wolkvox Studio.

In a Voice Flow, it can be used to build voice service experiences such as IVRs with menus, information capture and validation, transfers to skills, ASR, or TTS. In a Chat Flow, it makes it possible to create digital conversations capable of requesting information, evaluating responses, executing processes, taking different paths, and transferring the conversation when necessary.

What can Agentic Studio be used for in a Chat Flow?

Agentic Studio can help you build, among others, the following scenarios:

• Conversational bots for customer service, capable of requesting information and guiding the user toward an answer or solution.
• Data capture, such as name, email address, identification number, phone number, or other information required during the interaction.
• Response validation and evaluation, defining different paths depending on what the user writes.
• Transfer to human agents or skills, when the customer requests assistance or when a process condition requires it.
• Execution of operations or business logic, such as calculations, validations, or information processing.
• Use of functionality from other components available for the Routing Point, integrating it within the generated logic.
• Sending notifications or results, when the process requires communicating information obtained during the conversation.
• Design of closing, continuation, or exception paths, according to the detected intent or response.
• Rapid creation of conversational prototypes, which can later be tested, reviewed, and refined through new prompts.

What can Agentic Studio be used for in a Voice Flow?

In Voice Flow Routing Points, Agentic Studio can support the creation of IVRs with menus, information capture and confirmation, validation or error messages, timeout-based call termination, transfers to skills, and combinations of traditional logic with technologies such as ASR and TTS.

This makes it possible, for example, to design a service experience in which the user selects an option, provides their identification number, confirms the information, and is then transferred to the corresponding department.

• Accelerate IVR creation, reducing technical development time;
• Automate the writing of complex PHP logic;
• Reduce critical errors, especially syntax errors and session variable handling issues;
• Simplify the creation of menus, confirmations, and transfers;
• Support the migration of DTMF IVRs to more conversational experiences using ASR and TTS;
• Preserve the logic of key data from the beginning, such as identification numbers, contact variables, or service routes.

How to use the Agentic Studio component

Add Agentic Studio to the flow

1. Open the Routing Point you want to configure in wolkvox Studio. Within the “Cognitive” group, locate the “Agentic Studio” component, drag it onto the canvas, and connect it to the other corresponding elements in the flow. “Agentic Studio” can be part of a broader design or contain a significant portion of the logic within itself, since it can internally incorporate capabilities from other components available for that type of Routing Point. 
2. Double-click the component with the left mouse button to open its configuration panel.
3. Its configuration panel is described in detail below.

Write or dictate each prompt (instruction)

At the bottom, you will find a field where you can enter what you want to build or modify. You do not need to use programming syntax: describe the expected behavior using natural language.

For example, for a Chat Flow you could enter: I need a conversation that first asks for the customer’s name, then asks for two numbers, adds them together, and requests an email address to send the result. If the user asks to speak with a person, the conversation should be transferred to an agent queue.

Press “Enter” to send the instruction.

The response will appear within the conversation history.

If you prefer to explain the requirement verbally, use the microphone icon button located next to the input field.

When you start recording, the following elements will appear:

1. A duration counter and an audio waveform representing the audio being captured.
2. A button with a trash can icon that allows you to stop and discard the recording.
3. A Send button, represented by a paper airplane icon, to submit the audio.

When you send the recording, Agentic Studio transcribes the speech into text and uses that content as a new instruction within the conversation. The option to dictate the prompt is also part of the component’s functionality for voice flows.

You can freely combine written and recorded instructions during the same configuration process.

Review the history in the “Chat” tab

The “Chat” tab works as the conversation space between the administrator and Agentic Studio. In other words, it contains the conversation history.

You do not need to define the entire flow in a single instruction. Agentic Studio maintains the context of the conversation, so you can continue adding requirements, making corrections, or refining behaviors.

For example, after receiving an initial proposal, you can specify that the email address must be requested from the user, change the text of a response, add a transfer to a skill, or define what should happen when a response does not match any of the expected options.

Agentic Studio can also request additional information when it detects that a definition is required to correctly complete the logic. When designing voice flows, for example, it may be necessary to specify transfer destinations, confirmations, data to be captured, or behaviors that should occur after a selection.

This makes it possible to progressively shape the flow through a conversation instead of having to rebuild it manually every time a change is required.

While Agentic Studio processes the request, a loading indicator will be displayed. Once it finishes, you will also be able to see the time used to generate the response through an indicator such as “Worked for 3s”.

Review the generated flow in the “Code” tab

Once Agentic Studio has enough information to build the solution, you can open the “Code” tab.

For Chat Flow, this view displays a visual representation of the flow that Agentic Studio built internally. There you can identify the main steps, decisions, and possible conversation paths.

For example, the diagram may represent elements such as:

• Start of the conversation.
• Interaction with an AI conversational assistant.
• Evaluation of the user’s response or intent.
• Continuation of the conversation.
• Transfer to a human agent.
• End of the chat.
• Execution of functions or processes required to achieve the objective.

To the right of the diagram is the Diagram description, where a textual explanation of the generated solution is displayed. There you can review the summary, objective, and behavior of the different steps that make up the flow.

The labels displayed over the diagram paths make it possible to identify the conditions or intents that lead the conversation toward each result.

The result generated by Agentic Studio must be reviewed before being used in production. Automatic generation accelerates the design process, but the administrator must validate that the logic actually corresponds to the process they want to implement. This validation is also a fundamental consideration in the IVR-oriented version.

If, while reviewing the diagram or its description, you find something you want to change, return to the conversation area and enter a new instruction.

For example, you can request: Before asking for the numbers, ask for the user’s name.

Or: If the user asks to speak with an agent, transfer the conversation to skill 4011.

Or even request a broader modification to the logic that has already been created.

Agentic Studio will use the context of what has been built up to that point to update the proposal. After processing, return to Code and review the diagram again.

You can repeat this process as many times as necessary until you achieve the expected behavior.

Once you have reviewed the generated flow and are satisfied with the result, click “Save and Close”.

Then compile the Routing Point to correctly save and apply the changes made to the flow.

It is not recommended to finish the configuration immediately after receiving a response from Agentic Studio: first review the diagram and the generated logic, and then save and compile.

Review the generated result

1. Once the AI finishes building the proposal, you can review the result in the “Code” tab. There you will see the logic generated in PHP. In addition to the code, the component is designed to provide a clear structure that allows you to validate the flow behavior before using it.
2. Once you finish the interaction and review the result, click “Save and Close”.

Usage recommendations

For better results, it is recommended to:

• Write clear and specific prompts;
• Define the purpose of the flow from the beginning;
• Indicate what should happen in cases of errors, timeouts, or no input;
• Specify the desired voice if the flow will use TTS;
• Indicate the skill or destination to which the call should be transferred, when applicable.

Important considerations

Logic validation

Although Agentic Studio accelerates flow generation, reviewing the generated logic before considering it complete is mandatory. In particular, the pseudocode or expected flow behavior must be validated to confirm that it responds exactly to the business requirement.

Additional consumption

If the generated flow includes AI, ASR, or TTS components, this may generate additional credit or token consumption according to the current pricing.