The AI Assistant action runs when the workflow reaches its node. It can generate text, analyze a selected file, or call an AI Agent. To let a user trigger AI from an open form, use the AI Assistant form action step.
You can use the AI Assistant to:
Generate or refine text content during a workflow
Analyze images or documents uploaded in a form
Call an AI Agent within a workflow to execute intelligent business logic
Step 1: Open the Workflow Designer
Open an existing application, or create a new one.
Open the approval form designer. Prepare the fields or attachments that the workflow will use as input.
Click the Workflow tab at the top to open the workflow designer.
Step 2: Add the AI Assistant Action
In the left Actions panel, locate AI Assistant.
Drag the AI Assistant action onto the workflow canvas.
Connect the AI Assistant node after the steps that supply its inputs and before the steps that use its output.
Click the AI Assistant node on the canvas to open its configuration panel on the right.
All AI Assistant–related settings are configured in this panel.
Step 3: Configure Basic Information
Action name
Enter a name for the AI Assistant action. This name is used for identifying the node during workflow design and maintenance.
The action name does not affect the AI execution result.
Description
(Optional) Enter a description to explain the purpose of this AI Assistant action in the workflow.
Step 4: Select the Event type
The Event type defines how the AI Assistant is executed within the workflow.
The AI Assistant (Workflow Action) currently supports four Event types, displayed in the same order as the dropdown list. Each Event type determines the execution behavior of the AI Assistant and the configuration fields available in the settings panel.
Conversation with Assistant: Used for text-based conversations with the AI Assistant. This is the most general and commonly used AI invocation type.
Analyze image content: Used to analyze image content. An image file must be specified as input.
Analyze document content: Used to analyze document files (such as PDF or Word) and extract or understand their content.
Call AI agent: Used to call a configured AI Agent, allowing the Agent to execute more complex and reusable AI logic.
Select an event type to see the settings it needs. Choose the input source and the workflow variable that should receive the result.
1. Conversation with Assistant
When Conversation with Assistant is selected, the AI Assistant processes the input in a conversational, text-based manner.
Available settings:
User message
The main input sent to the AI Assistant. This is typically natural language text and can include workflow variables or form field values.Assistant instructions
Defines how the AI Assistant should respond, such as tone, format, or business context.Select input file(s) (Optional)
Allows you to select one or more files as additional input to provide more context for the AI Assistant.Including online research result (Optional)
When enabled, the AI Assistant may include online research results in its response.Save response to
Required. Specifies the workflow variable where the AI response will be stored.
2. Analyze image content
When Analyze image content is selected, the AI Assistant analyzes the specified image file.
Available settings:
Image
Select a value that supplies an image file, such as an image uploaded with the request.User message
Describes what information the AI should analyze or extract from the image.Assistant instructions
Defines how the analysis result should be returned, such as structured output or specific focus areas.Save response to
Required. Specifies the workflow variable where the image analysis result will be stored.
3. Analyze document content
When Analyze document content is selected, the AI Assistant analyzes the content of document files.
Available settings:
User message
Describes the analysis task to be performed on the document, such as summarization, information extraction, or evaluation.Assistant instructions
Defines the rules and output format for document analysis.Select input file(s)
Select one or more document files (such as PDF or Word) to be analyzed.Save response to
Required. Specifies the workflow variable where the document analysis result will be stored.
4. Call AI agent
When Call AI agent is selected, the AI Assistant calls a published AI Agent to perform a business task within the workflow.
Unlike general AI interactions, an AI Agent already contains its own persona, instructions, knowledge, tools, and response format. The AI Assistant action is responsible for selecting the AI Agent, providing the required business context, and saving the generated results for later use in the workflow.
The configuration includes the following sections:
Event type
Select Call AI agent to invoke a published AI Agent during workflow execution.
Application
Select the application where the target AI Agent is defined.
Only AI Agents published within the selected application are available for selection.
Select an AI agent
Choose the AI Agent that will process the current workflow action.
For example, an AI Agent can be used to:
Generate approval recommendations
Summarize workflow requests
Analyze support tickets
Classify business records
Generate workflow content
Produce structured business summaries
The AI Agent's behavior is fully determined by its own configuration, including its persona, instructions, knowledge, tools, and output format.
Include current context
When Call AI agent is selected, you can enable Include current context to automatically send workflow-related context to the selected AI Agent.
Instead of relying only on the input prompt, the AI Agent can generate responses based on the current workflow and its associated business data.
This is especially useful for business scenarios such as approval recommendations, ticket analysis, request summarization, intelligent routing, and workflow automation.
Available context
Current application context
Provides information about the current application and related business resources.
Workflow instance context
Provides information about the current workflow instance, including:
Workflow status
Request information
Applicant
Current process state
Workflow variables
Passes workflow variables that are available during the current workflow execution.
These variables may include business data entered by users or values generated by previous workflow actions.
Workflow tasks context
Provides task-related information, such as:
Current approval task
Assignees
Approval history
Task status
This allows the AI Agent to better understand where the workflow is currently running.
Tip
Enable the context the agent needs. In a test workflow, check which fields reach the agent before using the action with business data.
Input
Use the Input field to describe what you want the AI Agent to accomplish.
The Input defines the objective of the current action, while the selected context provides the business information needed to complete that task.
Output
Map the AI Agent response to a compatible workflow variable. If the agent exposes several outputs, map the ones later steps need.
For example:
Save response to
AIApprovalRecommendation
Once the response is stored in a workflow variable, it can be reused throughout the workflow, including:
Task forms
Approval pages
Notification emails
Later workflow actions
Other AI Assistant actions
Example
A common scenario is generating an AI approval recommendation before an approval task is assigned.
Workflow example:
Start ↓ AI Assistant ↓ Manager Approval
In this example, the AI Assistant executes before the Manager Approval task.
The agent uses the configured context to prepare a recommendation for the purchase request. The manager checks it against the request and supporting files before deciding.
The generated response can include:
Recommended action
Confidence level
Risk level
Key reasons
Missing information
Recommended next step
The response is saved to a workflow variable, allowing it to be displayed in approval forms, inserted into task notification emails, or reused by subsequent workflow actions.
Step 5: Save and Publish the Workflow
In a test workflow, check the normal result and the paths for missing input, failed analysis, and an empty response. Confirm that the next task receives the expected output or has a path for manual review.
Click Save in the top-right corner to save the workflow configuration.
Click Publish to apply the changes.
After publishing, run a test request and check the workflow instance. Verify that the AI node ran, the output variable was populated, and the next task received the result.
Summary
With the AI Assistant (Workflow Action), you can flexibly introduce AI capabilities into Yeeflow workflows for conversational content generation, image analysis, document understanding, or calling AI Agents to handle more complex business logic.
When configuring the AI Assistant, it is recommended to first choose the appropriate Event type based on your business goal, and then complete the corresponding settings to quickly build intelligent and automated workflows.






