Start Free trial

Articles in this section

See more →

Introduction to Wingz Workflows

Shweta Ajwani
  • Updated
Share:
close this to read article

This article covers the following:

Overview

A Wingz workflow automates repetitive analysis and reporting work that would otherwise require someone to manually pull data, review it, and act on it every time. Instead of having a team member regularly check on campaigns, generate reports, or turn raw data like heatmaps and session recordings into actionable next steps manually, you can set up a workflow that executes these tasks on its own (on a schedule or in response to a specific event). You can also configure the workflow such that it delivers the result directly to the relevant stakeholders. 

You can:

Both approaches produce the same kind of workflow and use the same building blocks described in this article.

A Wingz workflow builds this automation setup using three kinds of building blocks:

  • Start/End node - these are the start and end nodes that define when the workflow initiates and ends
  • Agent action - these are agent nodes that each run a predefined Wingz agent to perform a specific task, such as analyzing a heatmap or generating a PDF
  • Logic node - these are Condition, Duplicate, and Merge nodes that route, branch, or recombine the workflow as it runs

Once you connect these blocks and publish the workflow, a sequence of steps that would otherwise take manual, repeated effort runs automatically.

For example, consider a workflow that summarizes a paused A/B test. In this workflow:

  1. Its trigger fires whenever a campaign's status changes to Paused. 
  2. A single agent action reviews the paused test's performance and drafts a recommendation. 
  3. The workflow needs no branching, so no logic node is required.
  4. It ends by sending an email notification to the recipients.

The moment someone pauses a test, an email summarizing the results and next steps goes out, without anyone having to pull the report manually.

Note: Wingz workflow notifications are always sent by email.

The following sections define each of the building blocks you can use to build such workflows and how they fit together. The Sample Use Cases section later in this article covers several patterns you can adapt for your own team.

Launch the Workflows Canvas

The Workflows canvas is where you build and assemble a workflow.

To open the Workflows canvas:

  1. Log in to your Wingify account.
  2. From the left navigation panel, navigate to Wingz > Workflows. 
    This opens the Workflows dashboard, which displays a workflow template gallery above your workflow list, if you have any existing workflows.
  3. Open an existing workflow, click a template from the gallery, or click Create to start from a blank canvas.

For more information on choosing between a template and starting from scratch, see Create a Workflow. For the full list of available templates, see Explore Wingz Workflow Templates.

Understand the Workflows Canvas Layout

Once a workflow opens, regardless of how you started building it, the canvas is organized into four areas:

  • Wingz Workflow Assistant panel (left): a chat panel where you describe changes in plain language instead of placing nodes yourself. Click the collapse icon in the panel's header to hide it and free up space for the canvas, and click it again to bring the panel back. For more information, see Build a Wingz Workflow Using the Wingz Workflow Assistant.
     
  • Node palette (next to the canvas): a narrow strip of node icons you drag onto the canvas. Click the chevron at the bottom of the strip to expand it into labeled categories, including Core, Wingz, and Logic. To add a node, drag its icon from the palette onto the canvas, then release it where you want it placed.
     
  • Canvas (center): displays the workflow’s nodes and the connections between them. Drag from a node's edge to another node to connect them. Click a connecting wire, then press the Delete key to remove a connection.
     

  • Settings panel (right): shows the workflow's Name and Description by default. Click any node on the canvas to replace this with that node's own configuration options.

    Note: The settings panel always reflects your most recent selection. If you click away from a node without meaning to, click it again on the canvas to bring its configuration back in the panel.

Components of a Workflow

Every workflow is built from six node types, grouped into three categories. The table below summarizes the purpose of each node.

Category Category Description Available Nodes Purpose
Core Contains the trigger and end action every workflow needs. Start Defines when the workflow runs or initiates.
End Defines what happens after the workflow ends.
Wingz The AI agents that perform tasks. Agent Runs a Wingz agent that performs a task.
Logic The nodes that route, branch, or recombine the workflow. Condition Evaluates the value of the previous node in the workflow using an operator. The node routes the workflow down a true path or a false path based on the result.
Duplicate Splits one node's output into two or more parallel outputs.
Merge Combines the outputs of two or more nodes into a single flow.

This article further explains each node and its structure in detail. For more information on how to add and configure them, see Build a Workflow.

Start Node

The Start node determines when the workflow runs or initiates. You must choose and configure a Trigger Type to control how it's triggered. The available options are:

  • Time Based: runs the workflow on a recurring schedule.

    You must set the workflow run frequency depending on how often you want the workflow to run. You can choose to set the Schedule to Daily, Weekly, or Monthly.

    For example, a Time Based trigger set to Daily runs the workflow once every day, independent of anything else happening in the account.

    Note: Each schedule runs at a fixed time.

    • Daily runs at 7 AM EST every day
    • Weekly runs at 7 AM EST every Monday
    • Monthly runs at 7 AM EST on the first day of the month
       
  • Condition Based: runs the workflow when a campaign matches conditions you set. 

    You must configure the following trigger details:
    • Campaign Type: select the kind of campaign to watch, such as AB
    • Campaign Event: select the change that should trigger the workflow, such as Paused
    • Campaigns: select the specific campaigns the workflow must apply to

      For example, a Condition Based trigger set to Campaign Type: AB, Campaign Event: Paused, and Campaigns: Any runs the workflow whenever any A/B test in the account is paused.

Agent Node

The Agent node runs a Wingz agent, an automated task configured in Wingify with a specific goal. This node determines the action the workflow must take when it reaches this step, such as analyzing a heatmap or generating a PDF.

You must configure the following details for this node:

  • Select an agent: choose which agent the node must execute. Based on your requirement, you can choose from three options:
    • Select an existing agent from the list.
    • Select Wingz and add your task directly in the Additional Instruction field, which creates an agent from that instruction automatically.

    • Create a new agent to define a new agent's name, goal, and prompt from scratch within the node. This agent then becomes available in the dropdown list for selection.

      Tip: You can also create an agent directly through the Wingz Workflow Assistant chat, outside the node's own configuration panel. Prompt the chat assistant to ‘create an agent’ and describe what it should do. The assistant sets up the agent according to your instructions. Once created this way, the agent appears in the Select an agent list like any other, so you can reuse it across this workflow or future ones instead of recreating it each time.
  • Additional Instruction (optional): add context for this specific workflow run without changing the agent's configuration. What this field does depends on what you selected in the Select an agent dropdown:
    • If you selected an existing agent, Wingz passes your Additional Instruction to the AI alongside that agent's own Goal and Prompt. In this case, the instruction narrows or adjusts this one run without altering the agent itself. For example, if the agent's Goal is to analyze campaign performance, adding ‘focus on the mobile segment only’ scopes just this run to mobile data.
    • If you selected Wingz instead of an existing agent, this field defines the new agent's task instead of narrowing an existing one, since there's no prior Goal or Prompt for it to add context to.

End Node

The End node determines what happens once the workflow ends. You must choose and configure the following details:

  • Output Type to control this. The available options are:
    • Do Nothing: ends the workflow without any further action.
    • Notify: sends an email to a recipient you specify once the workflow ends.

If you select Notify, you must also configure the following under Notify me on:

  • Email: select this checkbox, then add recipients. Add each email address on its own line or separate multiple addresses with commas. Each address you add appears as a removable item above the address input box.

  • Generate AI email summary: select this checkbox to have Wingz summarize the workflow's results, including PDF reports, for the email recipients. Leave it unchecked to send the raw output instead. If the raw output exceeds the email's size limit, Wingify sends it truncated to that limit.

    Note: Selecting Generate AI email summary reveals an AI summary instructions field, pre-filled with a default structure that covers the outcome, key findings, recommendations, and caveats, written in Markdown-style formatting. Edit these instructions to change what the summary emphasizes, or click Reset to default to restore the original template. Wingz applies its standard safety rules to the summary automatically. You do not need to configure these separately.

For example, an End node set to Notify with Generate AI email summary checked sends a concise, AI-written summary every time the workflow ends. This is useful when the recipient just needs the takeaway and, if relevant, a PDF report. Leaving it unchecked instead sends the workflow's full raw output. This is useful when the recipient needs to review the underlying detail. An End node set to Do Nothing suits a workflow that only needs to update data in the background without alerting anyone.

Condition Node

The Condition node checks the output of the previous node in the chain. Based on the result, it routes the workflow down one of two paths. You must configure the following details for this node:

  • Operator: select how the previous node's output should be compared against the value you enter. Choose:
    • contains to match if the value appears anywhere in the output, 
    • == to match only on an exact match, 
    • != to match when the output is anything other than the value, or 
    • regex to match using a regular expression pattern
  • Value: enter the value to compare the previous node's output against.

The node produces two outputs based on the comparison:

  • True: taken when the comparison matches.
  • False: taken when it doesn't.

For example, suppose an Agent node's output includes the text ‘significant’ whenever a test reaches statistical significance. Connect a Condition node set to contains with the value significant after the Agent node, then route its two paths to different End nodes:

Condition Result Path Connect To

Outcome


 
Agent output contains ‘significant’ True Notify End node Sends the notification, since there's a result worth reporting.
Agent output does not contain ‘significant’ False Do Nothing End node Takes no action, since the result isn't significant.

This way, the workflow notifies stakeholders only when a test reaches statistical significance, and does nothing for every other result.

Duplicate Node

The Duplicate node splits one node's output into two or more parallel branches, so a workflow can run more than one Agent task simultaneously instead of sequentially. You must configure the following detail for this node:

  • Number of output nodes: set the number of parallel branches this node should split into. For example, set it to 3 to run three separate Agent tasks from this point in the workflow, such as generating three different test ideas from the same input.

    Note: The default is set to 2 and the maximum limit is 5.

Once you set the number of output nodes, the Duplicate node displays the corresponding output points on the node. Connect each one to a separate node, typically an Agent node, so every branch runs its own task independently of the others.

For example, a Duplicate node set to 2 output nodes can send one branch to an Agent node that analyzes a heatmap and another to an Agent node that analyzes session recordings, so both run at the same time.

Merge Node

The Merge node combines two or more parallel branches back into a single output, so the workflow can continue as one path again. 

Note: A Merge node waits for all connected branches to complete before it produces a combined output. If one branch is still running, the merge doesn't proceed until that branch finishes.

You must configure the following detail for this node:

  • Number of input nodes: set the number of parallel branches this node should combine back into one. This must match the number of branches you want to merge, for example set it to 3 to merge the results of three parallel Agent tasks into a single output.

    Note: The default is set to 2 and the maximum limit is 5.

Once configured, connect the output of each parallel branch into the Merge node, then connect its single output to the next node.

For example, a Merge node set to 2 inputs can combine the results of a heatmap analysis and a session recording analysis that ran in parallel, then pass the combined result to an End node set to Notify.

For more information on how to build workflows using the different nodes, see Build a Wingz Workflow Manually to Automate Tasks, or Build a Wingz Workflow Using the Wingz Workflow Assistant to build one conversationally. See Manage Wingz Workflows for information on how to find a published workflow in the list, edit it, and review its run history to confirm each node completed as expected.

Sample Use Cases

The patterns below show how the node types described above combine into complete workflows. Adapt the trigger, agents, and notification settings to match your own campaigns and reporting needs.

Use Case Workflow Structure Agent Instructions
Automated campaign summary Start → Agent → End (Notify) Analyze campaign results. Summarize each variation's performance, identify the winner, and list next steps.
Parallel campaign and UX analysis Start → Duplicate → [Agent 1 + Agent 2] → Merge → Agent 3 → End (Notify)

Agent 1: analyze campaign performance by segment.


 

Agent 2: summarize session recordings for friction points. 


 

Agent 3: combine findings into a PDF report.
Weekly experiment digest Start → Agent → End (Notify) List all running campaigns, their current status, traffic split, and any campaigns that have reached significance this week. Deliver as a summary.
Heatmap-to-test-ideas pipeline Start → Duplicate → [Agent 1 + Agent 2] → Merge → Agent 3 → Duplicate → [Agent 4 + Agent 5 + Agent 6] → Merge → End (Notify)

Agent 1: analyze session recordings.


 

Agent 2: analyze heatmap data. 


 

Agent 3: generate three A/B test ideas from the combined findings.


 

Agents 4 to 6: draft an implementation plan for each idea.
Paused-campaign debrief Start (Condition Based: Campaign Event = Paused) → Agent → End (Notify) Summarize the paused campaign's results, note whether it reached statistical significance, and recommend whether to relaunch, iterate, or archive it.
Conditional escalation Start → Agent → Condition → End (Notify) or End (Do Nothing) Evaluate the campaign against a target metric. Only route to the notifying End node when the result crosses the threshold, so routine runs stay silent.
Multi-stakeholder notification Start → Agent → Duplicate → End (Notify) + End (Notify) Summarize the campaign result once, then send the same summary to two different End nodes, each addressed to a different recipient, such as a product manager and a developer.

Need more help?

For more information or further assistance, contact Wingify Support.

Was this article helpful?

Return to top

Related articles