App Actions are custom transforms you create in Gladly to filter, reshape, and refine data from your connected systems for use in AI conversations. They live in the Custom Actions settings page under Admin → App Developer Tools.
App Actions are built on top of App Platform Actions, the read and write operations available from your organization's installed App Platform Apps. While App Platform Actions return data in a verbose, general-purpose format, App Actions let you transform that output into something specific and useful for a particular Conversation scenario.
For example, an App Platform Action might return a Customer's last 10 orders. An App Action could then filter that result to return only the single order a Customer is asking about, extract only the fields the AI needs, and make those values available for a response.
What App Actions are used for
Common uses for App Actions include:
Filtering data – Narrowing a large dataset down to a specific record (e.g., finding one order from a list of orders)
Removing sensitive information – Stripping PII or other sensitive fields before data is passed to the AI
Removing irrelevant data – Filtering out items that don't apply to the Conversation (e.g., removing digital products from an order)
Validating inputs – Checking whether a value meets certain criteria (e.g., validating a zip code)
Adjusting App Platform inputs – Modifying how data is passed into an App Platform Action to fit a specific use case
Accessing Customer context – Using built-in Gladly Customer profile data without needing an external API call
Key Concepts
There are four building blocks that make up every App Action:
Step – A single operation in your workflow. Steps execute sequentially, and each step can reference inputs and the results of previous steps.
Input – A value your App Action needs in order to run (for example, an order ID or Customer email). Inputs are defined by you and collected by the AI before the App Action executes.
Variable – An output value extracted from your App Action's final result and made available to the AI for use in a Conversation response.
Secret – A sensitive credential (such as an API key or token) stored securely and referenced in steps without exposing the value directly.