Skip to main content
Workflows in BotCadence allow you to create complex, stateful voice interactions. Instead of a single prompt handling the entire call, you can break down the conversation into logical nodes and paths, allowing for more controlled and robust voice agents.

​
Overview

A Workflow consists of nodes and paths. Each node can represent a specific state or sub-agent with its own specialized instructions. This is particularly useful for handling complex scenarios like:
  • verification flows (collecting ID, verifying date of birth),
  • support triage (identifying the issue catalog),
  • appointment scheduling.

​
Create a Workflow

To create a new workflow:
  1. Navigate to the Workflows tab in the sidebar.
  2. Click the + Create Workflow button in the top right corner.
  3. Your new workflow will appear in the list with a Draft status.
Workflow List

​
Configure a Workflow

Click on a workflow from the list to open the Visual Editor. The editor is divided into the Canvas (left) and the Configuration Panel (right).

​
Canvas

The canvas allows you to visualize the flow of the conversation.
  • Start Node: The entry point of the workflow.
  • New Path: Define different branches the conversation can take.
  • Nodes: Specific steps in the workflow, such as calling a sub-agent or performing an action.
  • End Node: Where the workflow concludes.

​
Configuration Panel

The configuration panel allows you to customize the behavior of the selected agent or node. It contains three main tabs:

​
1. General

Set the identity and core instructions for the sub-agent.
  • Agent Name: A descriptive name for the sub-agent.
  • Prompt / Instructions: The system prompt that guides the agent’s behavior for this specific step. Use this to define the agent’s goal and constraints for this part of the conversation.

​
2. Voice

Configure the speech and intelligence settings.
  • Conversation Model (LLM): Select the AI model (e.g., GPT-4o Mini) and adjust parameters like Temperature (Creative vs. Deterministic).
  • Speech Recognition (STT): Choose the provider (e.g., Deepgram) and the specific model for transcribing user audio.

​
3. Tools

Add capabilities to your agent, such as calendar integration or database lookups, allowing it to perform actions beyond just speaking.

​
Publishing

Once your workflow is designed and configured, click the Publish button to make it live. You can also Save your changes as a draft to return to later or check for Errors using the error indicator.