Automating Business Processes

This page explains how to automate business processes using Agentiqs workflow functionality. Workflows are defined by combining rules (conditions, branches, and so on) with AI agents. By fixing the steps with rules, omissions can be prevented, and by leaving only the processes that require judgment to AI, variation in responses can be reduced.

---

Creating a Workflow

This section explains how to create an Agentiqs workflow.

Procedure

1. Click [Workflow] > [New].
   
   • Creation options ([Create Manually] / [Create with AI]) will be displayed.
2. Select a creation method.
   • [Create Manually]: A blank workflow editing screen is displayed. Add steps manually.
   • [Create with AI]: Proceed to the screen where you enter what you want to create (see the next section for details).

Organizing Lists

In the list view, you can switch the display layout, group items by category, and pin items. For details, see "Organizing lists such as agents".

Creating with AI

This section explains how to tell AI what you want to build and have it generate a workflow automatically.

Procedure

1. Enter the details of the process you want to automate as the initial prompt.
2. Select the AI model to use.
3. Click [Start Creation].

tip

Building the workflow manually helps deepen your understanding of the content and enables an operation that continuously improves it. [How-to]-[Create a workflow for comprehensive review] explains the full creation process in detail. The following benefits are expected, so try creating one from start to finish:

• You will be able to understand the workflow output by AI, preventing it from becoming a black box.
• You can improve efficiency by dividing responsibilities, such as having AI create the overall structure while humans handle the details.

Editing the Workflow

This section explains how to edit a workflow you have created.

Procedure

1. Click [Workflow].
   • A list of workflows will be displayed.
2. Select the workflow you want to edit.
   • The workflow editing screen will be displayed.
   • For detailed parameter settings, see "Reference: Parameters per Step".
3. Click the [AI] tab in the right-hand pane.
   • The chat screen will be displayed.
4. Edit interactively with AI in chat and run tests as needed to check the results.

About workflows edited with AI

• Workflows created or edited interactively with AI do not always work correctly. Run a test to confirm that you get the intended results.

Define Inputs, Outputs, and Internal Variables

You can define workflow inputs and outputs, as well as internal variables that can be referenced from each workflow step. Define them from [Workflow Settings] by clicking the following items.

• Input/Output Parameters
  An input dialog is displayed when you run the workflow.

  
  If you set default values, you can save the effort of entering them.

  

• Internal Variables

  
  Defined internal variables

  Click {x} in each step's input field to display them as selectable options.

  

Saving and Restoring Workflow Versions

You can save workflow edits as versions and restore previous states as needed. Use this when you edit repeatedly in AI chat or when you want to preserve the state before making major changes.

Saving a Version

Click [Version] in the header of the workflow editing screen, then enter the version name and a description of the changes to save it.

AI will suggest a suitable version name and description based on the content.

The following actions allow AI to read the most recent edit differences and suggest a version name and description that match the content:

• In the AI chat on the workflow editing screen, instruct it with "Save this as a version."
  You can save the content you adjusted through dialogue as a milestone in your work.
• Click the button on the right side of the version name input field.
  

Restoring a Version

If you want to return to a previous state, select the version you want to restore on the version management screen and run [Restore].

Notes on Version History

• You can save up to 20 versions per workflow. When the limit is reached, older history is automatically overwritten.
• Version history is not carried over in the following cases:
  • When a workflow is duplicated or imported
  • When a workflow is moved between a shared-mode workspace and a non-shared-mode workspace

Executing a Workflow

This section explains how to run an Agentiqs workflow.

Procedure

1. Click [Workflow].
   • A list of workflows will be displayed.
2. Hover over the card of the workflow you want to run.
   • The [Run] button will be displayed.
3. Click the [Run] button.
   • If input variables are defined, an input dialog is displayed before execution.
   • The workflow switches to the [Execution History] tab and execution starts.

Viewing Execution History

On the workflow editing screen, you can not only create workflows but also review execution results and debug them on the same screen. In execution history, you can check variable values and execution logs in one place, and for steps that use AI you can also see execution time and cost.

• 1. Execution cost:
  • You can check the total cost for the entire execution and the equivalent amount in Japanese yen.
  • Clicking it displays a breakdown by step.
    You can identify AI steps or model settings with higher costs and compare the execution results before and after changes.
  
  Cost display applies only to steps where usage is recorded.
• 2. Variable pane, 3. Execution log pane
     On the tab, you can review each execution's status, start time, elapsed time, trigger name, and schedule name together, as shown below.
  • Execution date: Grouped by date for easier review
    
  • Title search: Quickly narrow down the execution you want to check.
    
  • Executions waiting for human review: A HIL badge is displayed.
    

Viewing Variables in the Workflow After Completion

You can check variable values after completion. You can reference the contents at the time execution completed, including variables that were not explicitly output.

The history list shows only the latest 50 entries per workflow. For older executions, you may not be able to view details such as step logs or final variables.

Running Automatically on a Schedule

Workflows that run repeatedly, such as periodic information gathering or aggregation, can be scheduled for automatic execution.

Procedure

1. Open the target workflow, then open [Schedule] in the workflow settings. (image ①)
2. Add a schedule and configure the following. (image ②)
   • Execution interval from [Hourly], [Daily], [Weekly], or [Custom]
   • Execution time
   Check the next scheduled execution shown on the screen. (image ③)
3. Enable the created schedule. (image ④)
4. After automatic execution, check the results in notifications, unread indicators, or [Execution History].

Operational notes

• Automatic execution runs in the environment of the user who enabled the schedule.
• Schedules received from a shared workspace are disabled. Only the user responsible for execution should enable them.

Debugging the Workflow

If a workflow does not behave as expected, use debug mode to isolate the step where the problem occurs. You can pause execution at breakpoints and run the workflow step by step while checking or changing variable values.

Procedure

1. Click the "Debug" button on the workflow editing screen.
   
2. Set breakpoints on the steps where you want to stop before execution.
3. Start a debug run.
4. When execution stops at a breakpoint, check the state.
   
   • 1. Check values in the variable pane and execution logs.
   • 2. Modify variables as needed, then select [Continue], [Step], [Rerun], or [Stop].
   • 3. To end debugging, click "End Debugging".
     End the session while debugging is stopped.

     Even if you exit debug mode, the workflow continues running.

Reference: Parameters per Step

The configuration parameters for each step, especially tool execution, require consideration of the interface and tool implementation, so they are expected to be configured through interaction with AI.

tip

See "Reference" for points to keep in mind when editing workflows.

Toolbox (Control Flow)

Loop

Processes each item in a list.

Configuration Parameters

Name	Required	Description
Iteration Target	Yes	Expression for the target to iterate over
Item Variable		Variable name that stores each item
Parallelism		Degree of parallelism (sequential when set to 1)
Maximum Executions		Maximum number of executions
Iteration Method		Options: "List (simple list)", "Tree (hierarchical structure)"
Child List Property Name		Property name of the child item list
Variable Name for Parent Element		Variable name that stores the parent element
Maximum Recursion Depth		Maximum recursion depth (unlimited if null)
Hierarchy Traversal Method (Depth-first/Breadth-first)		Options: "DepthFirst (depth-first)", "BreadthFirst (breadth-first)"

Conditional Branch

Branches processing based on conditions.

Configuration Parameters

Name	Required	Description
Condition Expression	Yes	Expression evaluated as a boolean

Parallel Execution

Runs multiple steps in parallel.

Configuration Parameters

Name	Required	Description
Parallelism		Maximum number of processes to run simultaneously

Sequential Execution

Runs multiple steps in order.

Configuration Parameters None

While Loop

Repeats processing while the condition is true.

Configuration Parameters

Name	Required	Description
Condition Expression	Yes	Loop continuation condition (repeat while true)

Toolbox (Integration and Connectivity)

HTTP Request

Sends an HTTP request to an external API.

Configuration Parameters

Name	Required	Description
Method	Yes	HTTP method
URL	Yes	Request destination URL
HTTP Headers		Dictionary of HTTP headers (JSON format)
Body		Request body (JSON format)
Store Result In		Dictionary of response destination variable names and values (JSON format)

User Inquiry

Queries the user during workflow execution and uses the answers in subsequent steps.
It supports approval, free text, single selection, and multiple selection.

Configuration Parameters

Name	Required	Description
Inquiry Type	Yes	Options: "Approve (Reject, Approve)", "FreeText (Text Input)", "SingleSelect (Single Selection)", "MultiSelect (Multiple Selection)" (※)
Inquiry Title		Title displayed in the inquiry dialog
Inquiry Description		Description displayed in the inquiry dialog
Output Variable	Yes	Workflow variable name where the result is stored
Initial Value		Default value or expression used as the initial value

※ About "Inquiry Type"

• You can specify the options as a fixed enumeration or as an expression that references a workflow variable.

  

• By using single selection or multiple selection to fix the response format, you can prevent conditional branch failures caused by variations in free-text input.

Toolbox (Data Processing)

Python Script

Executes Python code.

Configuration Parameters

Name	Required	Description
Python Code	Yes	Python code to execute
Timeout (seconds)		Execution timeout in seconds (1–3600; default is 1000 if unspecified)
Success Flag Variable	Yes	Variable name that stores execution success (true/false). If specified, the workflow continues even if an error occurs.

Toolbox (Basic Operations)

Set Variables

Sets values in variables.

Configuration Parameters

Name	Required	Description
Assignment	Yes	Dictionary of variables and values to assign (JSON format)
Variable Set Method		Options: "Overwrite", "Add"

Toolbox (AI Processing)

Call AI Agent

Calls a registered AI agent by name.

Configuration Parameters

Name	Required	Description
Agent Name	Yes	Name of the AI agent to call
AI Model Name		AI model (model ID)
System Prompt		System prompt that gives the model its role
User Message	Yes	User message sent to the agent
Store Result In		Variable name that stores the generated result

AI Conditional Branch

Branches based on a condition judged by AI.

Configuration Parameters

Name	Required	Description
AI Model Name		AI model (model ID)
System Prompt		System prompt sent to the LLM. If not specified, the base system prompt is used.
User Message	Yes	User message sent to the LLM

Knowledge Search

Searches a registered knowledge base.

Configuration Parameters

Name	Required	Description
Knowledge Name	Yes	Name of the knowledge base to search
Search Query	Yes	Query string used for the search
Maximum Results		Maximum number of results to retrieve (1–100)
Store Result In		Variable name that stores the search results

Call LLM

Runs a basic LLM generation.

Configuration Parameters

Name	Required	Description
AI Model Name		AI model (model ID)
System Prompt		System prompt that gives the model its role
User Message	Yes	User message sent to the LLM
Store Result In		Variable name that stores the generated result

Execute Tool

Runs a tool function.

Configuration Parameters

Name	Required	Description
Tool Name	Yes	Name of the tool to call
Function Name	Yes	Name of the function to execute
Parameters		Key/value parameters passed to the function
Store Result In		Variable name that stores the execution result
Success Flag Variable		Variable name that stores execution success (true/false). If specified, the workflow continues even if an error occurs.