The Workflow Builder Agent is currently in early alpha and available to selected customers. Contact your gaiia Customer Support representative if you'd like to request access.
Overview
The Workflow Builder Agent is gaiia's AI agent that helps you build workflows through conversation. Describe what you need in natural language, and the agent generates the workflow on the canvas in real time. You can refine through back-and-forth dialogue and edit manually at any point.
In this article, we'll be covering how to start a conversation with the agent, what kinds of requests work best, and how to refine workflows through conversation.
Start a conversation with the agent
The Workflow Builder Agent lives in a chat panel on the right side of the builder.
- Open or create a workflow in the builder.
- Click Ask gaiia in the builder toolbar.
- The agent chat panel opens on the right side.
- Type your request in the chat input (for example, "Create a workflow that sends a welcome email when an account is created").
- Press Send or hit Enter.
- Watch the canvas as the agent builds the workflow live.
- Continue the conversation to refine the workflow.
The agent writes nodes to the canvas in real time — you see progress immediately! You can edit the canvas manually at any point, even while the agent is still working.
What to say to the agent
The agent responds to plain-language descriptions of what you want the workflow to do. Be specific about triggers, actions, and conditions.
Initial requests
Describe the trigger and the actions you want the workflow to perform:
- "When a ticket is created, get the account details and send a notification to the assigned technician"
- "When an account status changes to active, if the account type is residential, create a welcome work order"
- "Create a manual workflow that calls the provisioning API, waits for success, then updates the account status"
Follow-up requests (conversational editing)
Once the agent has built an initial workflow, you can ask it to modify specific parts:
- "Add error handling to the email node"
- "Change the condition to check if account type is commercial"
- "Add a Slack notification when the workflow fails"
- "Remove the wait node"
The agent validates every workflow it generates against gaiia's workflow schema before showing it on the canvas. If validation fails after retries, the agent reports an error and you can rephrase your request or edit manually.
While the agent is working
The chat panel shows a thinking animation while the agent generates the workflow. The canvas updates in real time as nodes are added.
You can interact with the workflow while the agent is working:
- Edit nodes manually — add, delete, or modify nodes while the agent is still building; your changes won't conflict with the agent's updates
- Draft your next message — you can start drafting a new request before the agent finishes the current one
- Collapse the chat panel — the agent continues working in the background
Review and refine the workflow
Once the agent finishes building or modifying the workflow:
- Review the workflow on the canvas.
- Test the workflow using the Test button to validate behavior.
- Continue the conversation to request changes, or edit nodes manually.
- Click Publish when the workflow is ready to go live.
Your conversation history is preserved per workflow — you can return later and pick up where you left off.
Tips for working with the agent
- Describe what should happen — the agent understands business logic, so you don't need to be overly prescriptive. You can say "send a welcome email when an account is created" rather than trying to specify node types or technical details
- Start simple and refine — build the basic flow first, then ask the agent to add conditions, error handling, or notifications in follow-up messages
- Ask for explanations — if you're unsure what the agent built, ask it to explain specific parts of the workflow
- Treat mistakes as part of the process — if something isn't right, describe what needs to change and the agent will adjust. You don't need to start over
Related to