AI Configurations Overview

Summary

What are AI Configurations?

AI Configurations give you one dedicated space in Akeneo PIM to manage all your AI content-generation settings. You can create, edit, and test your AI setups from a single central page under Settings → AI Configurations.

This unified approach ensures that your enrichment logic is consistent across products, Channels, and Locales, while reducing duplication and configuration effort. It lays the foundation for scalable, automated product content creation powered by AI.

AI Configurations Grid

The AI Configurations grid in Akeneo PIM, listing existing configurations with their label, code, type, target Attribute, and enabled status.

Default model vs. Bring Your Own Key (BYOK) add-on

Every Akeneo PIM environment includes an Akeneo-managed AI model (currently GPT 4.1 Mini) that powers all AI Configurations out of the box — no setup required.

With the optional Bring Your Own Key (BYOK) paid add-on, you can go further by choosing a specific AI model from a supported provider and connecting it with your own API key. This gives you greater control over model capabilities, cost management, and compliance.

  Default (included) BYOLLM add-on (paid)
AI model Akeneo-managed (GPT 4.1 Mini) Choose from supported OpenAI or Anthropic models
API key Not required — managed by Akeneo You provide your own API key from the selected provider
Cost Included in your plan (fair-usage policy) Billed to you directly by the AI provider
Model selection No selection possible Per-Configuration selector in the edit form

The BYOK add-on is a paid feature. If you do not have it, the model selector appears locked in the AI Configuration form. Contact your Customer Success Manager to enable it.

 

Supported providers and models

Provider Model Best suited for
Anthropic Claude Opus 4.6 Complex reasoning, nuanced content, detailed analysis
Claude Sonnet 4.6 Balanced performance for everyday enrichment and translation
OpenAI GPT 4.1 High-quality generation with strong instruction-following
GPT 4.1 Mini Cost-effective for high-volume tasks
GPT 5 Advanced reasoning, complex multi-step tasks
GPT 5.4 Latest generation with improved accuracy
GPT 5.4 Mini Cost-effective latest-generation alternative

The list of available models may evolve over time as Akeneo adds support for new releases from each provider.

 

For details on selecting a model and entering your API key, see Creating & Managing AI Configurations. For security and data handling, see AI Configuration Security & Data Privacy.

Configuration types

Each AI Configuration in Akeneo PIM belongs to one of three types, which determines how the AI processes your data:

Type What it does
Generation Generates new content for a target Attribute using product data, assets, and a prompt as input. Use this for descriptions, SEO metadata, summaries, and similar enrichment tasks.
Translation Translates the value of an Attribute from a source Locale to one or more target Locales. Does not require a target or source Attribute — it relies on source and target Locale selection instead.
Asset Auto-Tagging Automatically analyzes image and document assets and generates descriptive tags or labels. Configured separately through the asset management flow.

Choosing the right type at creation time defines which configuration fields are available and how the AI Configuration behaves when executed.

Accessing the AI Configurations page

In Akeneo PIM, go to Settings → AI Configurations to access the AI Configurations page.

You need the appropriate permission to access Settings → AI Configurations. If the menu item is not visible, contact your PIM administrator.

From the AI Configurations page, you can:

  • Browse all existing AI Configurations
  • Create new configurations from scratch
  • Edit, duplicate, enable, or disable existing ones
  • Test configurations using the built-in Preview tool

AI Configurations Page

The AI Configurations list view in Akeneo PIM, showing configuration labels, types, target Attributes, and their enabled/disabled status.

Structure of an AI Configuration

Each AI Configuration acts as a reusable blueprint that defines how AI should generate or translate content. A configuration contains the following sections:

1. General information

Basic identifiers for your configuration:

  • Label: A human-readable name for the configuration (for example, "SEO Meta Generator").
  • Code: A unique internal reference used in automation rules. Cannot contain spaces.
  • Configuration Type: Choose between Generation, Translation, or Asset Auto-Tagging.
  • Status: Enable or disable the configuration.
  • Enable for All Attributes: When enabled, the configuration is not restricted to a single target Attribute and can be applied to any supported Attribute type.

2. Target & source

Defines what data the AI uses and where the generated output goes. You can select:

  • A specific Target Attribute where generated content is written (for example, Product Description).
  • One or more source Attributes that provide input context (for example, Product Name, Material, Short Description).

Target and source Attribute selection does not apply to Translation-type configurations. Translation configurations use source and target Locales instead.

3. PX Insights user reviews (optional)

If PX Insights is enabled for your environment, you can choose to include relevant customer reviews automatically as part of the AI input context. This additional data can help the AI emphasize authentic product benefits or highlight highly-rated features.

Contact your Customer Success Manager for PX Insights activation details.

4. Assets (optional)

Include image or document assets as part of the context used by Akeneo AI. Assets can come from Asset collection Attributes in the PIM. This allows the AI to analyze visuals and extract details — such as product color, materials, or certifications from technical datasheets.

Supported asset types include JPG, PNG, WEBP, TIFF, and PDF (first 20 pages). For full details on supported formats and quantity limits, see AI Configuration Considerations & Limitations.

5. Instruction prompt

The instruction prompt defines how the AI processes your selected data. In an AI Configuration, you can:

  • Choose from predefined Akeneo prompt templates for common use cases.
  • Write a custom instruction to achieve a specific goal.

Well-designed prompts lead to better, more consistent results. Use the AI Prompt Optimizer and Rewriter to evaluate and improve your prompts after creating a configuration.

Channel-level and Family-level prompts

In addition to the base prompt defined in an AI Configuration, you can add generation instructions at the Channel level (via the Generative AI tab on each Channel) or the Family level. At generation time, the final prompt is built by combining the AI Configuration base prompt, Family-level instructions (if defined), and Channel-level instructions (if defined). This lets you maintain a single base configuration while tailoring outputs for specific destinations without duplicating configurations.

AI Configurations Complete Config

A fully configured AI Configuration in Akeneo PIM, showing the General Information, Target & Source, Assets, and Instruction Prompt sections.

Benefits of using AI Configurations

Consistency across your catalog

When prompts are centralized in a single AI Configuration, every product enriched by Akeneo AI follows the same tone, format, and structure — helping maintain a cohesive brand voice at scale.

Dynamic brand guidelines (tone of voice)

You can scale a single AI Configuration across a multi-brand catalog by linking products to Reference Entities (such as Brands or Collections). Akeneo AI automatically fetches Attributes from the linked Reference Entity record — such as Tone of Voice rules or glossaries — and applies them to the generation context based on the selected Channel and Locale.

Supported Reference Entity Attribute types for AI context injection include Text Area, Simple Select, Multi Select, and Number.

Time savings and scalability

Once validated, a single AI Configuration can enrich hundreds or thousands of products automatically through the Rules Engine or through manual execution in the Product Edit Form using the Ask AI panel.

Easier testing and iteration

Use the integrated Preview tool to test an AI Configuration on specific products before saving or applying it broadly. You can preview both Generation and Translation outputs instantly.

Reduced maintenance

If a prompt or enrichment strategy changes, updating it in one AI Configuration automatically applies the update everywhere that configuration is used — in manual runs and in automated Rules Engine rules alike.

Next steps