Smart Variants Agent - User Guide

This User Guide details the end-to-end process of using the Smart Variants Agent to transform existing Bill of Materials (BOM) configurations into structured Variant Management models within Aras Innovator.

The Smart Variants Agent is a purpose-built Task Assistant Agent within the InnovatorEdge AI framework.
It enriches the digital thread by analyzing the configuration relationships within a product’s 150% BOM data to infer rules and automate the creation of governed Variant Management items.

The agent facilitates the configuration process by performing the following:

• Direct System Access: Reads BOM structures directly from Aras Innovator via the Edge API, requiring no manual document uploads.
• Rule Extraction: Sends a BOM representation to the AI Agents Service to derive variant rules through deterministic co-occurrence analysis.
• Intelligent Visualization: Displays a merged 150% BOM, coupled (AND) rules, either/or (XOR) rules, and a dependency matrix (Implies/Exclude).
• Governed Commitment: Creates Variant Management entities in Innovator only after the user executes the final import.

• Manual Reduction: Reduces the engineering effort required to manually model variability from BOM data.
• Verification Separation: Separates the analysis phase from the import phase to ensure mandatory human validation before any data is committed to Innovator.
• Digital Thread Integration: Supports downstream configuration management with structured Variant Management objects.

The quality of the AI-driven analysis is directly proportional to the integrity of the source data in Aras Innovator.
For optimal results, ensure:

• Multiple Configurations: The product must have multiple 100% BOM configurations.
  The agent compares these to determine which components are mandatory, optional, coupled, or mutually exclusive. Single-configuration products will not produce meaningful rules.
• Standardized Naming: Parts with identical keyed_name values across configurations are identified as the same component.
  Inconsistent naming can result in duplicate or missing rules.
• Structural Depth: The agent retrieves and analyzes structures up to nine nesting levels of Part_BOM navigation properties.
  Structures exceeding this limit may not be fully represented.
• Clean Data: Ensure relationships in the system of record are clearly defined, as missing or ambiguous links reduce the accuracy of inferred rules.

The workflow begins in the Select Product interface, where users identify the target for configuration modeling.

1. Identify Target Product: Use the Search products field to filter the list by name, item number, or description.

2. Select Target: Click a row in the table to highlight a single product.

   

3. Click Analyze: Selecting Analyze in the top-right header initiates the process.

4. Monitor Progress: A non-dismissible Analyzing modal tracks progress as the agent fetches the BOM expansion and generates configuration rules.

   

Once analysis is complete, the agent moves to the Review step to present the Analysis Results.
This environment allows domain experts to verify the AI’s interpretation before system commitment.

The interface summarizes findings in a single line (e.g., “1 root item in 150% model • 14 coupled rules • 4 either/or rules • 345 dependency rules”) and presents detailed data in four functional tabs:

• 150% BOM: An expanded tree view showing the merged hierarchy across all configurations. Mandatory items are distinguished from optional items.

• Coupled Rules: A searchable table of parts that always appear together across analyzed configurations.

• Either/Or Rules: A view identifying mutually exclusive groups where exactly one option is valid per configuration.

• Dependency Rules: An interactive matrix mapping Implies (if A, then B) and Exclude (A and B cannot coexist) relationships.

Use the interactive toolbar to Expand All or Collapse All groups during auditing.

After clicking Continue in the review phase, the process moves to the final commitment step.

This screen allows you to configure the final import parameters:

1. Review Entities to Create: A summary panel lists the counts of entities to be generated, including Variability Items, Features, Options, Variable Components, Usage Conditions, and Asset Links.

2. Define Identifiers: Enter an Item Number Prefix (e.g., VM-eBike).
   This is prepended to all generated item numbers.

   

3. Click Create in Innovator: This initiates the Import Progress modal, which executes in three phases:

• Phase 1: Core vm_VariabilityItem (including features and options).
• Phase 2: vm_Usage conditions.
• Phase 3: vm_VariableComponent and asset relationships.
  

Upon successful commitment, an Import Summary is displayed.

• Open in Innovator: Click the provided link to open the Root Variability Item in the native Aras Innovator interface.
• System Records: In Innovator, you can navigate the resulting Variability Structure and Usage Conditions tabs to verify the generated model.
• Continuous Workflow: Select Import Another to return to the selection step for a different product.
  

Once the model is committed to Aras Innovator, it becomes part of the authoritative digital thread.
You can perform the following directly within the native interface:

• Metadata Review: View the generated Number, Revision, and State (defaulting to Preliminary).
  

Variability Items: shows a flat list for auditing all child entities, displaying their system-generated numbers, descriptions, and current lifecycle states.
You can select individual entries to perform native Innovator actions such as edits or lifecycle promotions.

• Variability Structure: Navigate the Variability Structure tab to see the multi-level hierarchy of items and features established by the agent.

All items created by the agent are subject to standard Aras Innovator Governance.
This means:

• User Permissions: Your ability to see, edit, or delete the generated Variant Management items is strictly tied to your native Aras Innovator identity and assigned roles (e.g., Configuration Management Group).
• Lifecycle Management: Committed items follow native lifecycle transitions (e.g., Preliminary to In Review) and versioning rules established in your environment.

• Ensure Multiple Configurations: The agent identifies mandatory, optional, or mutually exclusive components by comparing different 100% BOMs.
  Products with only a single configuration will not produce meaningful rules.
• Standardize Part Naming: Components must use consistent keyed_name values across all variants to be identified as the same item.
  Inconsistent naming results in duplicate or missing configuration rules.
• Adhere to Input Constraints: To ensure reliable processing and retrieval, keep search field inputs under 100 characters and item number prefixes to a maximum of 12 characters.
• Human-in-the-Loop Validation: All AI-derived rules and the 150% BOM must be reviewed and confirmed by an engineer before committing the model to the system of record.
• Maintain Active Session State: Do not close the browser window or navigate away while the Analyzing or Import Progress modals are active to prevent the loss of session data.