Introduction

Once you understand what AI Configurations are, the next step is to create your own. Each configuration defines how Akeneo’s AI engine should use your product data, assets, and instructions to generate or translate content.

By setting up reusable configurations, you ensure consistency in tone, format, and data usage across your catalog. You can then apply these configurations manually or automatically through the Rules Engine.

When to Create an AI Configuration

Create a new AI Configuration when you need to:

• Automate repetitive enrichment tasks (for example, generating product descriptions, SEO metadata, or technical specs).
• Standardize tone and format for multiple product families.
• Translate existing content into new locales with consistent style.
• Perform data checks or transformations (for example, summarizing key attributes or validating consistency).
• Leverage assets (like images or PDFs) to extract additional context.
• Each configuration can serve a different purpose, but all share the same underlying structure.

How to Create a New Configuration

Follow these steps to create your first configuration:

Step 1: Open the AI Configurations Page

Go to Settings → AI Configurations.

• You’ll see a list of all existing configurations and their status.

Click Create Configuration in the top-right corner.

Step 2: Complete the General Information Section

In the General Information panel, provide the key identifiers for your configuration:

Once you create your configuration, you can begin to create your prompt immediately, and offer more context for your needs! 

Step 3: Define the Target and Source

This section defines where AI should write results (target) and which attributes it should use as context (source).

Target Attribute:

• Select the field where the generated content should be stored (for example, Description, SEO Meta Description, or Material).
  This determines the output type (text, select value, etc.).

Source Attributes (Data to take into account):

• Select the product attributes that the AI should analyze to create the new content.
  Example: For a SEO Meta Generator, you might include Product Name, Short Description, and Unique Selling Points.

If you want the configuration to be generic and work across multiple attributes, you can apply it to “All Attributes”. This will help you use that configuration on any supported attribute type.

Step 4: (Optional) Add PX Insights User Reviews

If your Akeneo instance has PX Insights enabled, you can choose to include relevant customer reviews automatically.

Including reviews allows AI to emphasize real customer feedback or highlight top-rated features.
This can be especially useful for creating benefit-driven content or summarizing user experiences.

Step 5: (Optional) Include Assets

Assets allow AI to analyze images or documents related to a product.

In the Assets section, you can:

• Include all assets from your product record.
• Limit the scope to specific asset collections (for example, “Main Images,” “Certificates,” or “Datasheets”).
• Choose to include Media Files or Media Links.

This step is optional but valuable for use cases such as:

• Detecting dominant product color from images.
• Extracting certifications from PDF documentation.
• Enriching descriptions using visual attributes.

Step 6: Write or Select an Instruction Prompt

The Instruction Prompt tells AI exactly what to do with the provided data.

You can:

• Select a predefined prompt from Akeneo’s library for common use cases.
• Write your own custom prompt to achieve a specific goal.

When writing a prompt:

• Clearly define the AI’s role (e.g., “Act as an SEO copywriter”).
• Specify the task (e.g., “Generate a short meta description under 155 characters”).
• Include any formatting or tone requirements.
• Provide examples if necessary.

Once written, you can test and optimize your prompt later using the AI Prompt Optimizer and Rewriter. 

Before saving the prompt you can also utilise the Preview feature to help you see the response from AI. 

Saving and Managing Configurations

When you finish filling in the details, click Save. Your configuration will now appear in the list view with its label, target attributes, type of configuration, and status.

From the list view, you can:

Example AI Configurations

Example: “SEO Meta Generator”

Target Attribute: SEO Meta Description

Source Attributes: Product Name, Short Description, USP

Assets: Packshot

Prompt: 

Act as an SEO copywriter for e-commerce. Using the available attributes (name, brand, short_description, Product_Type, supplier), write a single-sentence meta description.

Requirements:
- Punctuation and spaces count toward the limit. No emojis or special characters beyond standard punctuation.
- Tone: concise, persuasive, and trustworthy; avoid boilerplate.
- Data mapping and priority:
  1) Use product name and brand (if brand missing, omit).
  2) Add one clear key benefit drawn from short_description. If unclear, infer a likely benefit from Product_Type.
  3) End with a direct call to action.
- Include brand only when present; do not fabricate data; do not include supplier unless it is the consumer-facing brand.
- Validation: output only the meta description text with no quotes, labels, or extra text.
- If insufficient data to form a clear benefit, default to a general value-oriented benefit (e.g., “designed for reliable performance”) and still include a CTA.

This configuration can then be used:

• In the Product Edit Form for manual generation
• Or through the Rules Engine for scheduled automation

Next Steps

Once your configuration is created, you can:

• Test it immediately using the Preview feature to ensure the output meets your expectations.
• Refine or rewrite your prompt using the AI Prompt Optimizer and Rewriter for better results.
• Apply the configuration at scale using the Rules Engine.

Continue to AI Prompt Optimizer and Rewriter to learn how to check your configuration's prompt and review generated content before saving or automating it.