Note: During the beta phase, automation actions using OpenAI will run on Frontify’s API key—no need to use your own. You can process up to 10,000 assets per month. In the future, you’ll need to use your own OpenAI key. Frontify’s key may only be available for limited use cases, such as trials. |
This is a general introduction to automations. If you’d like to automatically set asset descriptions, check out this article instead. |
Table of contents.
Intro
Frontify automations are rules you set up to make things happen on their own when something changes, like when a file is uploaded or edited. You can:
Choose what starts the automation - the trigger (like uploading an image),
Set conditions (like “only if it’s a PNG”),
Decide what action should happen (like “add a description” or “send a message”).
Install
Frontify’s core automation actions, like setting asset workflow status, descriptions, titles, and tags, are installed by default. To use OpenAI-powered actions such as generating descriptions, analyzing images, or translating metadata, go to the Marketplace and, under the app section, search for 'OpenAI automation actions' to install them (procedure here).
You can access automation from your project or library's header navigation. (Note: In the old navigation, access was from the power bar on the left-hand side.)
With the old UI, the journey to arrive at the automation dashboard is a little different, as shown in the GIF below.
Set up new automations
Once in the automations page, you can set up a new automation using the “New automation” button.
When you click the “New automation” button, a new automation flow will appear in draft status. Before activating it, you need to configure the flow according to the automation needs. Automations don't run until they're activated.
As a good practice, set a meaningful name for the automation flow after a new draft is created.
The automation setup screen is where triggers, trigger conditions, and actions are configured. Click the “New automation” button to see the automation setup screen.
Trigger configuration
After clicking on the “Add a trigger” button, available triggers are displayed. You can select one trigger from this list:
Trigger configuration - Conditions
After the trigger is specified, you can optionally set some conditions on the trigger, as the image below shows.
Action configuration
After the trigger is configured, it is time to configure the action(s). After clicking on the “Add an action” button, you can see the available actions. Depending on what the automation is expected to do, you can select one of the actions from this list.
It's important to understand the logical sequence of actions: If an action (e.g, Add tags to the asset) is using the result of another action (Analyze image with OpenAI), then they should be in the correct order in the flow. For example, the "Add tags to the asset" action should not be placed before the "Analyze image with OpenAI" action.
Action configuration - input parameters
Once the action is selected, you have to configure it based on your needs. Each action accepts different input parameters. Input parameters are values that the automations need to perform an action. Typically, input parameters can come from one of the following:
The trigger
One of the previous actions
It is hard-coded
Some examples are as shown below:
For the “Set asset title” action, the “Id of the asset” comes from the trigger (asset ID), and the value for the “Title to be set” comes from one of the previous actions (action 54).
For the “Translate with OpenAI” action, the value of the “Text to be translated” is coming from one of the previous actions, and for “Target language” a hard-coded value (“German”) is set by the user.
It's also possible to combine a few of these values to pass to a parameter.
Activate the automation
Once all actions are configured, you can activate the automation via the toggle in the top right corner.
To activate the automation, ensure that at least one trigger and one action are selected, and that all required parameter fields for each action are filled out. Otherwise, the automation cannot be activated.
Active automations will automatically run when the corresponding trigger event occurs and execute the configured actions.
Limitations and FAQs
Frontify Automations are designed to simplify and scale repetitive tasks — but like any powerful tool, there are a few things to keep in mind. We’re actively working to improve these areas and will continue to evolve the experience based on feedback.
What states of automations are there?
What states of automations are there?
Draft and active. There is no versioning of the automations.
Can I use automations across libaries or projects?
Can I use automations across libaries or projects?
Currently, automations work within a single library or project.
Can I rename action steps to make them easier to reference?
Can I rename action steps to make them easier to reference?
Not just yet — but this improvement is on our roadmap. Currently, action steps are labeled with internal placeholders (like action-55), which can make them a bit tricky to recognize when building more complex flows. We’re exploring ways to make these identifiers more visible and easier to manage in the UI.
Why don’t automation steps show in the asset’s audit trail?
Why don’t automation steps show in the asset’s audit trail?
At the moment, automation actions are triggered via the public API, but they’re not reflected in the asset audit trail. We're investigating this and plan to surface these events in the future to provide better traceability and transparency.
Why didn’t my automation run?
Why didn’t my automation run?
Several factors can affect whether an automation is triggered successfully. Here are the most common causes:
File format compatibility
Some AI-based actions (like image analysis) only support specific file types. Check the top of individual action help articles to confirm supported formats.
Timeouts
In some cases, automations — especially image-related ones — may time out due to system slowness or external rate limits (like OpenAI’s).
Delays in custom metadata
When used at scale, we apply internal delays to prevent conflicts. This can cause slower performance in large batch operations.
Strict filtering from AzureOpenAI
If you’re using our newer AI image analysis actions, these may occasionally reject content that doesn’t meet AzureOpenAI’s safety filters.
We're making improvements to provide clearer error messages and more visibility when something goes wrong — so you can troubleshoot (or ask Support to!) more effectively.
Do automations retry automatically if they fail or time out?
Do automations retry automatically if they fail or time out?
Currently, they don’t. If a timeout occurs, the automation run is considered complete. We plan to offer clearer status visibility and may introduce manual re-run options for failed or skipped assets in the future.
Are there any other known considerations?
Are there any other known considerations?
Here are a few things to watch out for:
Certain file formats (e.g., PSDs or non-standard video files) may not work reliably with specific automation actions.
Image-based actions are more prone to timeouts, especially during peak system activity.
Bulk custom metadata operations may run more slowly due to safeguards against data conflicts.
AI-based image actions using AzureOpenAI may be subject to stricter moderation rules.
We’re working toward better status indicators — so you’ll know whether each step succeeded, failed, or was skipped, and why.
And that’s a wrap — thanks for reading. If you have a specific use case in mind, check out our guide on setting up automatic asset descriptions for step-by-step instructions.