You have a bunch of tests, right? With every build of your application under test, you run them all. If a test fails, application should be fixed, or you test should be fixed. Wouldn't it be nice to make how-to documents from that tests?
Time for an example (Ruby, Watir):
You run that test, and you get output like this (this is real output from my test):
I posted this at wtr-general list and people recommended to take a look at rSpec and Systir.