Using Generative AI (With or Without AI Configurations)

Summary

Overview

Once you have created and tested your AI Configurations, the next step is to execute them, either manually on a single product or automatically across your catalog.

Akeneo provides two ways to run AI Configurations:

  1. Manual execution from the Product Edit Form, ideal for testing or one-off enrichment.
  2. Automated execution through the Rules Engine, ideal for recurring or large-scale enrichment.

Both methods use the same configuration logic, ensuring that the AI behaves consistently no matter how it is triggered.

Manual Execution via Product Edit Form

Manual execution allows you to generate AI-powered content directly from within a product’s record.
This is ideal for:

  • Testing a configuration on a specific product
  • Making small adjustments before full automation
  • Enriching new or priority products before a campaign launch

Ask AI Options

From the Product Edit Form, click Ask AI to choose how you want to generate or translate content. This panel offers two quick actions (no setup required) and two configuration-based actions (for more control).

GenAI - Options in the PEF

Smart Fill

Automatically generates a suggested value for the selected attribute. You can use this feature on its own for quick, one-off enrichment, which simply uses the product information that is already filled in. Alternatively, you can select an AI Configuration within Smart Fill when you need more control over how the content is produced (for example, applying specific guidance, tone of voice, or Reference Entity data).

GenAI - Generation without an AI Configuration

Smart Translate

Translates an existing attribute value from one locale to another. You can use this on its own for a fast, standard translation. Alternatively, you can select an AI Configuration within Smart Translate when you need more advanced translation behavior, such as respecting specific brand glossaries or formatting instructions.

GenAI - Translation With an AI Configuration

 

How to Execute Manually

  1. Open a product from your Product Grid.
  2. Click the Ask AI button in the top right corner.
  3. Choose Generate from Configuration if you already have created a configuration, or Smart Fill if you want to quickly generate a value without any prior configuration.
  4. Choose the attribute you want to enrich.
  5. Select the desired AI Configuration from the dropdown list.
  6. Click the AI Configuration of choice.

The generated value will appear directly in the attribute field. You can review and edit it before saving the product.

Step-by-step Guide

  • Click Ask AI in the top right corner.

 

  • Choose Generate from Configuration if you already have created a configuration, or Smart Fill if you want to quickly generate a value without any prior configuration.

 

  • Choose the attribute you want to enrich.

 

  • Select the desired AI Configuration from the dropdown list.

 

  • Click Fill the Attribute and receive a result
 
 

 

Supported Output Attributes

You can run AI generation on the following attribute types:

  • Text
  • Text Area
  • Simple Select
  • Multi Select
  • Reference Entity Single Link
  • Reference Entity Multiple Links
  • Number
  • Measurement
  • Yes/No (Boolean)
  • Date

When to Use Manual Execution

Use manual execution when:

  • You’re testing a new prompt before automating.
  • You want to enrich only a few attributes in a product (via the Product Edit Form).
  • You need full control over reviewing content before saving.
     

Automated Execution via Rules Engine

For large-scale or recurring enrichment, AI Configurations can be triggered automatically using the Rules Engine.
This enables teams to generate, translate, or audit content for thousands of products at once, without manual effort.

How to Execute via Rules Engine

  1. Go to Settings → Rules.
  2. Create a new rule or open an existing one.
  3. Define conditions that identify which products should be affected.
  4. Example: All products in the “Shoes” family that have an empty SEO meta description.
  5. Under Actions, select Generate attribute value with AI.
  6. Choose the desired AI Configuration from the list.
  7. Save and enable the rule.

Once the rule runs, the AI Configuration will automatically generate or update attribute values for all products that meet the rule’s conditions.

Note: If appropriate filters are not applied in the rule, AI configurations can override existing content present in the product

 

 

Monitoring Automation Performance

Automation can process thousands of records per execution. To ensure smooth performance:

Factor Description
Scope The number of products meeting rule conditions affects total runtime.
Attributes used More context attributes = longer processing.
Locales Each locale adds a separate generation task.
Assets Large or numerous assets (images, PDFs) increase processing time.

This varies based on data size, attribute types, and concurrent system activity.

 

Tips for Reliable Execution

Validate in Preview First
Always test your configuration on a few sample products using the Preview tool.

Start with Small Rules
Run smaller batches first to verify results before scaling to your entire catalog.

Use Descriptive Rule Names
Example: generate_missing_descriptions_fr or audit_incomplete_materials.

Keep Rules Simple
Avoid combining too many actions or conditions in a single rule for clarity and reliability.

 

Example Workflow

Here’s how a typical team might operationalize AI Configurations:

Step Action Outcome
1 Create AI Configuration (“SEO Meta Generator”) Defines prompt, attributes, and tone
2 Test using Preview Verify quality of generated output
3 Create Rule in Rules Engine Define target scope (e.g., missing SEO descriptions)
4 Enable Rule AI starts generating automatically when rules run
5 Review & Monitor Track enriched data and iterate prompt as needed

 

Troubleshooting

Issue Possible Cause Suggested Fix
AI generation did not run Configuration disabled or not selected in rule Ensure configuration is enabled and linked correctly.
Output appears inconsistent Prompt may be too broad or unclear Refine the prompt using the AI Prompt Optimizer.
Rule runs too slowly Scope too large or asset-heavy products Limit rule scope or optimize asset size.
Nothing appears in target attribute Attribute type mismatch Verify configuration relevance and attribute compatibility.

 

Best Practices Summary

  • Use manual execution for validation and targeted enrichment.
  • Use Rules Engine automation for continuous, large-scale execution.
  • Regularly analyze and rewrite your prompts for accuracy and tone consistency.
  • Keep automation monitored and adjust scope for optimal system performance.

Next Steps

After setting up automated or manual execution: