Workflows

Workflows are the execution strategy for your chatbot conversations. Build complex conversation flows using a visual drag-and-drop interface.

What is a Workflow?

A workflow defines how your chatbot processes and responds to user messages. It consists of multiple nodes connected together, where each node performs a specific action in the conversation flow.

Every workflow must have a Start Node (entry point) and an End Node(exit point). Between these, you can add any combination of processing nodes to create sophisticated conversation experiences.

Creating a Workflow

  1. Navigate to Studio → Workflows
  2. Click Create Workflow
  3. Enter a descriptive name for your workflow
  4. You'll see a canvas with Start and End nodes already connected
  5. Click the + button on connections to add new nodes
  6. Drag nodes to rearrange the visual layout
  7. Click on any node to configure its settings

Node Types

Each node type serves a specific purpose in your conversation flow:

Start Node
The entry point for every conversation. Configure auto-reply messages and initial instructions that greet users when they start chatting.
Agent Node
The AI-powered brain of your chatbot. Configure LLM providers, attach tools and knowledge bases, define response models, and customize prompt behavior.
Condition Node
Control conversation flow with conditional logic. Define rules using AND/OR operators based on previous node responses and route to different paths.
Form Node
Render custom forms to collect structured data from users. Link forms you've created in the Forms section and capture user submissions.
Static Message Node
Send fixed, pre-defined messages. Include dynamic variables from previous node responses to personalize the content.
Action Node
Perform external actions: send emails, post Slack messages, make API calls, or create Zendesk tickets. Supports retry and error handling.
End Node
The exit point of your workflow. Attach widgets here to deploy your chatbot and configure sharing options.

Best Practices

Start Simple

Begin with a basic Start → Agent → End flow. Add complexity gradually as you understand how nodes interact.

Use Descriptive Names

Name your workflows and nodes clearly (e.g., "Lead Qualification Bot", "Check Intent") to make them easy to understand and maintain.

Test Incrementally

Use the live preview to test your workflow after adding each new node. This helps identify issues early.

Handle Edge Cases

Use Condition nodes to handle different user intents and unexpected inputs gracefully.

Next Steps

Agent Node Deep Dive

Learn about LLM configuration, tools, and knowledge bases

Tools

Explore the IT Services Intent Classifier and other tools