Automations
Automate catalog enrichment with continuous monitoring
Automations generate suggestions automatically when changes happen in your catalog. Anything you can do manually - generating descriptions, filling in attributes, making bulk edits - you can automate.
Set up triggers to run automations when specific events occur, like an attribute being updated or an asset being connected. Enable auto-publish and changes go live automatically without review.
Why use automations
Automations eliminate repetitive work. Set up an automation once instead of manually creating proposals every time products need updating.
They're especially valuable for new products. For example, set up an automation to enrich empty descriptions - new products from your data source get enriched automatically as they sync.
Turn on auto-publish for trusted enrichments and changes go live without review. Focus on complex work while automations handle routine maintenance.
How automations work
Automations run when their triggers fire. You define one or more triggers - when any of them fires on an item that matches your filters, the automation creates suggestions just like manual proposals. The suggestions work identically - AI generates content, you review, then publish.
Triggers
Triggers define when an automation runs. There are two trigger types:
- Attribute updated - Runs when a specific attribute changes on an item. For example, trigger when a product's description is updated to automatically generate a matching meta description.
- Asset connected - Runs when an asset (image, video, etc.) is connected to an item. For example, trigger when a product image is added to automatically generate alt text.
You can add multiple triggers to an automation. The automation runs when any of its triggers fire.
Choosing triggers
Pick triggers based on what data the automation needs. If your automation generates meta descriptions from product descriptions, use an "Attribute updated" trigger on the Description attribute. If it generates alt text from images, use an "Asset connected" trigger.
Each item gets enriched once per automation to prevent infinite loops. To re-enrich items, create a new automation.
Creating an automation
Navigate to Automations from the side menu and click Create automation.
New to proposals?
Automations use the same filters, attribute selection, and generation settings as manual proposals. If you're new to this workflow, check out Generating with AI first to understand the basics.
Naming and setting catalog filters
Give your automation a descriptive name that explains what it does. "Enrich empty meta titles" is clearer than "Meta automation" when you're managing multiple automations.
Choose which items to monitor using the same filtering options as catalog views. Select the entity type (products, variants, bundles, etc.), then add filters for categories, brands, attributes, or any other criteria.
Be specific with filters to target exactly what needs enrichment. "Products where Meta Title is empty" is better than "All products" - it avoids processing items that don't need work.
Testing your filters
Before creating the automation, use the same filters in your catalog view to see what will match. This helps verify you're targeting the right items.
Setting up triggers
After configuring filters, choose what triggers the automation. Click Add trigger and select a trigger type:
- Attribute updated - Pick which attribute should trigger the automation. For example, select "Description" to run whenever a product's description changes.
- Asset connected - Triggers whenever an image or other asset is connected to a matching item.
Add as many triggers as you need. The automation runs when any trigger fires on an item that matches your filters.
Parent entity triggers
For variants and SKUs, you can also trigger on attributes from parent entities. A variant automation can trigger when a product-level attribute changes, and a SKU automation can trigger on product or variant attributes.
Choosing attributes and strategy
Select which attributes to enrich and which languages to generate. You can choose between AI generation (with custom instructions) or manual edits (find/replace, set values).
This works exactly like creating a manual proposal. Configure generation instructions, enable deep research if needed, set character limits - all the same options apply.
Attribute instructions
Automations use the generation instructions configured on each attribute. Update attribute settings to change how automations generate content for that attribute.
Enabling on creation
Choose whether to enable the automation immediately or keep it disabled until you're ready. The status section shows how many attributes will be enriched upon creation if existing items already match your filters.
If you're testing settings, leave it disabled, create it, then enable once you've verified the configuration is correct.
Auto-publish
Auto-publish makes automations fully hands-off. Suggestions publish automatically to your connected systems without review.
Toggle auto-publish from the automations list or automation detail page anytime.
When to use auto-publish
Auto-publish works best for simple, predictable enrichments where you trust the output. Examples:
- Filling empty meta titles with product names
- Standardizing formats (adding units, fixing capitalization)
- Translating content from a source language
- Setting default values based on product attributes
- Copying information between fields
These tasks have clear, objective rules. The output is predictable, and errors are unlikely.
When to be careful with auto-publish
Complex scenarios can work with auto-publish but need careful setup:
- Marketing descriptions requiring brand voice
- First-time AI generations with new instructions
- Content using deep research on products with limited online information
- Attributes with strict business rules or compliance requirements
Configure these properly as you would for manual proposals. Test without auto-publish first, review several batches for consistency, then enable auto-publish once confident.
Test before auto-publishing
When setting up a new automation, run it without auto-publish first. Review a few suggestions to verify the output matches expectations, then enable auto-publish once you're confident.
Managing automations
The automations page shows all automations with their status, triggers, pending suggestions, last run time, and who created them.
Click any automation to view its generated suggestions. To enable or disable an automation or toggle auto-publish, click the ellipsis menu on the right side of the row.
Reviewing suggestions
Suggestions from automations appear in the proposals section. They work exactly like manual proposals - click through items, review generated content, make edits if needed, and publish.
If auto-publish is enabled, suggestions publish automatically and you'll see them marked as published when you view the automation.
Enabling and disabling
Disable an automation to pause it. It stops checking for new matches but keeps existing suggestions. Enable it again to resume.
Disabling doesn't affect existing suggestions
Disabling doesn't delete or revert suggestions already created. They remain as normal proposals.
When to create a new automation
You can't edit an automation's filters, triggers, or attributes after creation. To change what it targets, what triggers it, or what it enriches, create a new automation.
Important limitations
Each item gets enriched once per automation to prevent infinite loops. To re-enrich items, create a new automation.
You can't edit an automation's triggers after creation. To change triggers, create a new automation.
Common use cases
| Use Case | Trigger | Filter Example |
|---|---|---|
| Generate meta descriptions from descriptions | Description updated | "Products where Meta Description is empty" |
| Generate alt text when images are added | Asset connected | "Products where Alt Text is empty" |
| Translate updated content | Description updated | "Products where Swedish Description is empty" |
| Enrich new products from data source | Name updated | "Products where Description is empty" |
| Fill missing meta titles | Description updated | "Products where Meta Title is empty" |
| Add technical specs | Name updated | "Products in Family: Electronics where Specifications is empty" |
Best practices
Test without auto-publish first. Review a few suggestions, then enable auto-publish once you trust the output.
Use specific filters. "Products where Meta Title is empty" only processes items needing work, avoiding unnecessary re-processing.
Choose meaningful triggers. Pick triggers that correspond to when the automation's input data changes. If your automation generates meta descriptions from product descriptions, trigger on the Description attribute - not on unrelated attributes.
One automation, one job. Separate automations for descriptions, meta titles, and translations are easier to manage and let you control auto-publish individually.
Monitor pending suggestions regularly if auto-publish is off. Check the automations page periodically for items waiting review.
Plan carefully. Items process once per automation, so test your filters and settings manually before automating.