Translation actions allow AI Workflows to automatically translate product content into other languages.

Instead of manually creating separate translation jobs, translations can now become part of a larger automated workflow pipeline.

Translation actions can run independently or after other workflow actions such as:

This allows workflows to automatically generate and translate content in sequence.

Translation actions analyze existing product content and generate translated versions for selected fields.

Examples:

Translations are generated using AI models configured inside the workflow.

Managing multilingual catalogs manually is often:

AI Workflows automate this process by integrating translations directly into enrichment pipelines.

This helps:

Translation workflows are commonly used for:

A webshop enriches Dutch product content first and then translates it automatically into German.

Workflow:

This entire process runs inside one AI Workflow.

A Translation action consists of:

Products are processed individually through the translation pipeline.

Inside a workflow:

You will then enter the Translation configuration screen.

The Translation configuration screen allows you to define how translations should behave.

Common settings include:

You can select:

Example:

Different models may affect:

You must define which fields should be translated.

Examples:

Only selected fields are processed.

Prompts define how the AI should translate content.

Simple example:
"Translate the Shopping Description to German."

Advanced prompts may include:

Example:
"Translate the Shopping Description to German using professional and natural grammar suitable for ecommerce product pages."

Translation actions often perform best after enrichment actions.

Example:

This creates much stronger multilingual outputs than translating raw supplier data directly.

The scope determines which products the translation action processes.

This is often controlled through:

Focused translation scopes improve:

Before running large translation actions, you can test translations on selected products.

Testing helps validate:

Testing is strongly recommended before large scale execution.

Test results display:

You can:

Reasoning explains why the AI generated certain translation choices.

Example:
"German ecommerce terminology was used to improve readability and product discoverability."

Reasoning improves moderation transparency.

Confidence scores indicate how certain the AI is about the translation quality.

Higher confidence often means:

Lower confidence may indicate:

After testing:

Products are then processed individually through the translation pipeline.

Results become visible inside the Results tab.

The Results tab displays:

This helps monitor translation execution.

Depending on the workflow configuration:

Once approved, translated content synchronizes back into the product data.

Translated content quality improves significantly when:

Avoid translating poor quality supplier content directly when possible.

Good prompts specify:

This creates more natural translations.

Different languages may require:

Separate workflows improve maintainability.

Low confidence outputs may require:

Confidence scoring helps prioritize moderation.

Example:
A webshop wants to automate multilingual Google Shopping content.

Workflow:

This creates an automated multilingual enrichment pipeline.