Skip to main content

Create your first AI Workflow

Create your first AI Workflow in Elovate by setting up a trigger, adding enrichment and translation actions, testing the output, and synchronizing approved content.

Create your first AI Workflow by combining a trigger with one or more AI actions.

In this example, you will:

  • detect products missing shopping descriptions

  • generate enriched content

  • translate content to German

  • review output before synchronization

Step 1: Open AI Workflows

In Elovate:

Enter a workflow name.

Example: German Shopping Enrichment

Step 2: Configure the trigger

The trigger determines which products enter the workflow.

Example trigger rules:

  • Category equals Dry Cat Food

  • Shopping Description is Empty

  • Status equals Enabled

Only matching products will be processed.

Step 3: Add a Content Enrichment action

Click: Add new action → Content Enrichment

Setup


Action name: Enriching shopping title + Description

AI platform: OpenAI (or Gemini or Claude)

AI model: the recommended model Elovate preselects (for OpenAI currently GPT-5.6 Luna)

Enable moderation: Yes

  • Editors: your name

  • Moderators: your name

What attributes do you want to fill:

Fields:

  • Shopping name

  • Shopping description

Example prompt:

Write an SEO-friendly shopping description in Dutch for {{name}} by {{brand}}, based on {{description}}. Professional e-commerce tone focused on pet nutrition. Maximum 500 characters. Only use facts from the product data.

The placeholders ({{name}}, {{brand}}, {{description}}) give the AI the product data. Without them, the AI has nothing to write from. See Using prompts in AI Workflows.

Step 4: Test the action

Before running the workflow:

  • Click Test products

  • Select a few products

  • Review the generated output

Check:

  • content quality

  • formatting

  • tone of voice

  • confidence scores

Testing helps prevent large scale workflow mistakes. A test is a preview: nothing is synced to your webshop.

Step 5: Create a translation setting

Before adding a Translation action, make sure a translation setting exists.

To create one:

  • Go to General Settings

  • Open Settings

  • Go to Translations

  • Create a translation setting with the correct source and target language

Example:

Admin → Pip's - DE

The translation setting defines which language pair the workflow should use. Its source store must be the same as the workflow's source store.

Step 6: Add a Translation action

Return to your workflow and click:

Add new action → Translation

Configure:

  • Action name

  • Moderation settings

  • Translation setting

  • Translation provider

  • Glossaries

  • Attributes to translate

Step 7: Configure the Translation action

Select the translation setting you created earlier.

Then choose whether the action should:

  • override existing content

  • translate each item only once

  • use a glossary

  • translate specific attributes only

For the translation provider, you can select DeepL or an AI provider.

DeepL translates automatically and ignores prompts.

AI providers (OpenAI, Gemini or Claude) allow a custom prompt per attribute. Use it only for tone and terminology; the languages and glossary terms are added automatically.

Step 8: Choose attributes to translate

Select the fields this action should translate.

Example:

  • Shopping Title

  • Shopping Description

  • Product Description

  • SEO Title

  • Meta Description

Only selected attributes will be translated.

Step 9: Configure the action scope

In the Scope section, choose which products the action should run on.

You can run the action on:

  • all products in the current workflow scope

  • a limited set of items

For testing, choose a few test items first.

You need at least one test item before previewing the output.

Step 10: Save and test the Translation action

Click:

Save & test prompt

Review the translated output before running the workflow.

Note: a translation test translates the text as it is in Elovate now, not the test output of the enrichment action. Use it to judge tone and terminology.

Check:

  • grammar

  • terminology

  • formatting

  • ecommerce tone of voice

  • confidence scores

Step 11: Run the workflow

After testing your actions, return to the workflow overview and run the workflow.

The workflow processes matching products through each action in sequence. A product only moves on to the next action after its result has been approved and synced.

Example:

Trigger → Content Enrichment → Translation

Step 12: Review generated output

Open the Results tab to review:

  • generated descriptions

  • translated content

  • confidence scores

  • AI reasoning

  • moderation states

Approve correct output and decline incorrect results if moderation is enabled.

Step 13: Synchronize approved content

After approval, synchronized output updates your product data automatically.

Example platforms:

  • Magento

  • webshop catalogs

  • marketplace feeds

Example workflow structure

Trigger → Content Enrichment → Moderation → Synchronization → Translation → Moderation → Synchronization

Did this answer your question?