I don't know how good of an idea it is, but in my Agile GUI Testing software, one thing I'm working on (debugging my old code, actually) is some very basic image recognition software. If/when it works, it will let the tester write a line of code like this to find an image on screen:
(x,y) = find_image_click_coordinates 'moose-image.jpg'
This gives you the x/y coordinates of the "center" of the image. In this example, I'm looking for the image of the two moose that are currently shown on the front page of my OneMansAlaska.com website.
In the past I've found this to be very useful, because then you can tell the software, "Keep looking at the screen, but wait until you see this image, and then do <fill-in-the-blank>." For instance, if you're waiting for Acrobat to be displayed, you can have a small Acrobat image that you use, and just tell the system to wait until it sees that image.
Of course you can do other things that are system-specific to determine if Acrobat is displayed, but at the moment I like this approach because a tester can implement this very quickly and easily.
June 1, 2010: Our Automated GUI testing software now has a new home page. Just follow that link for downloads and support documents for our Automated/Agile GUI Testing software.