Workflow

Workflow is a visual builder that lets you create and manage AI-powered conversation flows. Instead of handling every user question in the same way, you can connect different steps on a canvas and define how the chatbot should respond to each situation.

A workflow is made up of nodes connected by paths. When a user starts a conversation, Curatu follows the connected path one step at a time until the workflow reaches a point where the conversation can be completed.

Workflow can be connected to supported channels such as Web, WhatsApp, Microsoft Teams, and Voice.

What You Can Do with Workflow

Workflow helps you build conversational experiences that go beyond a simple question-and-answer setup.

You can use a workflow to:

• Answer questions using a Knowledge Base.
• Use an Agent to generate responses and perform specific tasks.
• Ask users for information and use their answers in later steps.
• Create conditions to send users through different paths.
• Search across multiple Knowledge Bases and use the most relevant one.
• Connect external tools and APIs to an Agent.
• Create structured responses using Chat Canvas.
• End a conversation with a custom message.

The type of workflow you create depends on the experience you want to provide to your users.

How a Workflow Works

A workflow is built by placing nodes on a canvas and connecting them together.

Every workflow starts with a Start node. You then add the nodes needed for your conversation and connect them in the order you want them to run.

For example:

Start → Knowledge Base

A more advanced workflow might look like:

Start → User Feedback → If / Else → Agent

When a user interacts with your chatbot, the workflow follows the connected path based on the configuration of each node.

A workflow follows one path at a time. Different nodes can determine what happens next based on the user’s message, previous results, or other conditions.

Workflow Nodes

Nodes are the individual building blocks of a workflow. Each node performs a specific task.

Curatu provides nodes for different types of conversation logic and actions.

Core Nodes

Core nodes provide the main building blocks of a workflow:

• Start — The point where every workflow execution begins.
• Agent — Uses an AI agent to perform a task or generate a response.
• Auto Orchestrator — Selects the most relevant Knowledge Base to provide the most relevant answer.
• End — Explicitly ends the workflow with an optional final message.

Logic Nodes

Logic nodes help you control what happens next in the conversation.

• User Feedback — Asks the user a question and uses their response in the workflow.
• If / Else — Checks conditions and sends the conversation through the appropriate path.

Tool Nodes

Tool nodes provide access to specific sources of information.

• Knowledge Base — Searches stored content and uses it to answer the user’s question.

Build a Workflow

To create a workflow:

1. Open Build → Workflow.
2. Select Create.
3. Add the nodes you need to the canvas.
4. Connect the nodes to define the conversation flow.
5. Configure each node.
6. Test the workflow.
7. Publish the workflow when it is ready.
8. Deploy the published version when you want to make it live.

The canvas automatically saves changes as you build your workflow, so there is no separate Save action.

Connect Nodes

Connections determine which step runs after the current step.

For example:

Start → Knowledge Base → Agent

In this flow:

1. The workflow starts.
2. The Knowledge Base searches for relevant information.
3. The Agent handles the next step.

Some nodes have specific connection rules. For example, Auto Orchestrator can connect only to Knowledge Base nodes.

Test a Workflow

Before making a workflow live, use Test mode to try the conversation yourself.

You can:

• Start a new chat.
• Review previous test conversations.
• Open chat sessions.
• Debug individual messages.
• Inspect execution details.
• Trace the path followed by the workflow.

When debugging a message, you can review information such as the input, output, error details, execution time, and credits used.

Workflow Versions

Workflow versions let you keep a published version separate from your current draft.

This is useful when you want to make changes without immediately changing the version that is being used by your connected channels.

The basic workflow lifecycle is:

Draft → Published → Deployed

Workflow uses three main statuses:

Publish

When you publish your changes, Curatu creates a new version from the current draft.

You can choose to deploy the new version immediately or publish it without deploying it.

If you publish without deploying, connected channels continue using the previously deployed version.

Deploy

Deploy a published version when you want connected channels to start using that version.

This lets you prepare and publish changes without immediately changing the live experience.

Connect a Workflow to a Channel

After your workflow is ready, you can connect it to a supported channel.

Use Connect to Channel from the workflow menu to connect the workflow to:

• Web
• WhatsApp
• Microsoft Teams
• Voice

Workflows used with Web channels use the V2 chatbot experience.

A workflow must be deployed before a connected channel can use that version.

Share a Workflow

You can share a workflow with other members of your team so they can collaborate with you.

When sharing a workflow, you can give a person one of the following access levels:

• Editor — Can edit and share the workflow.
• Viewer — Can view and test the workflow.

The Owner has full access to the workflow, including editing, sharing, publishing, deploying, renaming, and deleting it.

Credits

Workflow displays an estimated credit usage for a workflow run.

The actual number of credits used can depend on which nodes are executed and how those nodes are configured.

Only nodes that are executed during a run consume credits.

The credit amount shown on the canvas is an estimate. Actual usage can vary depending on the path taken during a conversation.

Workflow Access

Workflow is available on the Enterprise plan.

The following roles can access Workflow:

• Super Admin
• AI Bot Analyst

Additional permissions apply to specific workflow actions.

Common Workflow Patterns

You can use different node combinations depending on what you want your chatbot to do.

What’s Next

Once you understand the basic workflow structure, you can learn how to configure each node and build more advanced conversation flows.

• Start — Configure conversation context and incoming-message guardrails.
• Knowledge Base — Answer questions using your stored content.
• Agent — Build AI-powered tasks and responses.
• Auto Orchestrator — Route questions across multiple Knowledge Bases.
• User Feedback — Collect information directly from users.
• If / Else — Create conditional conversation paths.
• End — Close a workflow with a final message.