Skip to content

AI Employee Actions

An AI Employee job is a sequence of actions - individual steps that the AI Employee performs one after another on its own computer. Each action does one specific thing: focus a window, type text, click a button, read a file, update a process field, and so on. By chaining actions together, you can automate complete pieces of desktop and browser work.

Actions run in order, from top to bottom. Every action reports either success or failure, and you decide what happens next in each case: proceed to the next action, jump to another action, repeat the action, or exit the job. Many actions also support a retry count, so a step that depends on a slow screen can quietly try again until it succeeds or runs out of attempts.

Each action has its own settings (for example, the Text to Type for a typing action, or the Window Name Contains for a window action). Two features make these settings dynamic:

  • {{Field}} tokens - most text settings accept {{Field Name}} placeholders. When the job runs, the placeholder is replaced with the value of that field from the process that launched the job. For example, a typing action configured with {{Invoice Number}} types the actual invoice number for the current process.
  • Action output tokens - many actions save a result as they run (the text retrieved from a page, the value on the clipboard, the name of the window that was found). Later actions can reference these saved values by inserting the corresponding token into their settings with the Insert Field Token button.

The action catalog is organized into the following categories:

For how to create the jobs that contain these actions, see Managing AI Employee Jobs.