Having said that, it seems that Automator could use conditionals and branching execution paths. Something like "troll through a bunch of news sites and pick out all the pictures with their captions...now, if the caption includes the word 'Bush', add a red overlay to the picture, but if it includes the word 'Kerry', add a blue one...now, move all the modified pictures into my iPhoto library and put them in an album with today's date".

Considering the layout of Automator's window, it's easy to imagine that two "steps" could exist side-by-side, below a conditional step (one input, two or more potential outputs). Another step below could take the results of multiple inputs and pull them back together to continue the workflow.

Perhaps it's too complicated for a simplified interface. But I'm pretty sure I'd want this kind of flexibility before I'd created my third real task flow in Automator. Am I alone?