Workflow actions are the AI powered tasks executed after a trigger is activated.

Actions are the core processing layer inside AI Workflows. They define what should happen once products enter the workflow.

Examples include:

A workflow can contain one or multiple actions connected together.

Once a product matches the workflow trigger:

This creates a sequential automation pipeline.

Actions execute in sequence from left to right.

Each action can build on the output of the previous action.

Example:

In this example:

This chaining behavior is one of the most powerful parts of AI Workflows.

Previously, enrichment and translation often required separate manual jobs.

Workflow actions combine these processes into a single automated pipeline.

Example workflow:

Everything runs inside one workflow structure.

AI Workflows currently support several action types.

Attribute Extraction analyzes existing content and fills structured attributes automatically.

Examples:

Useful when supplier content lacks structured attribute data.

Content Enrichment generates or improves product content using AI.

Examples:

This helps improve catalog quality and discoverability.

Translation actions translate content into other languages automatically.

Examples:

Translation workflows can be chained after enrichment workflows.

Category Mapping helps assign products to categories using AI.

Useful for:

Content Validation reviews content quality and identifies missing or problematic data.

Examples:

Quality Scoring evaluates how complete or reliable product content is.

This can help prioritize products needing manual attention.

Each action contains its own settings and configuration.

Depending on the action type, configuration options may include:

This allows workflows to be highly customizable.

Before running an action on large product groups, actions can be tested on selected products.

Testing helps validate:

Test results may include:

Testing is strongly recommended before large scale execution.

Actions can contain moderation steps before synchronization.

Depending on the configuration:

This improves quality control for AI generated content.

After configuration and testing, actions can be executed manually or automatically depending on the workflow type.

When an action runs:

Large workflows may take time depending on:

Actions can contain multiple execution states.

Examples:

Statuses help monitor workflow progress and identify bottlenecks.

Avoid creating large complex workflows immediately.

Start with:

Then expand gradually.

Focused workflows are easier to maintain.

Good examples:

Avoid combining unrelated objectives into one large workflow.

Always validate:

before running workflows on large catalogs.

Not every workflow requires manual approval.

Examples:
High moderation:

Low moderation:

Example:
A webshop imports incomplete supplier products.

Workflow:

This entire process runs automatically through connected workflow actions.