To create an automation task there's the Web Recorder, which monitors your browser's interactions with a Web site and creates a task that can be edited and run. The only situation where this recording technology doesn't work is when a Web page contains ActiveX controls. Then you have to manually add to the script specific mouse actions (e.g. "go to these coordinates in this window and click").
A future release will include detection of interactions with ActiveX controls and will automatically add the mouse actions to your task. You can also create tasks using the Object Recorder, which understands the broader range of Windows controls. This lets you create tasks that will automate any application.
An interesting feature of the task editor is the ability to record thumbnails of browser and application screens when a task is run, which provides a visual record that makes modifying a task weeks or months later somewhat easier.
There's also a Workflow Editor for the construction of sequences of tasks with conditional logic and a Report Designer so that you can track the creation of tasks as well as task and workflow runs.
I mentioned AAS's ROI tracking. This feature keeps track of a range of factors including task development and debug time, documentation time, training, and the execution times of tasks allowing AAS to report on what the tasks would have cost if humans had done them – a neat feature when it comes to justifying investment.