Jonathan Kohl blogs about the dangers of reckless test automation. He's right on the money and I strongly recommend reading what he has to say.
One thing I'd like to highlight is that the source of the problem is not "100% test automation" itself but the interpretation of it. I haven't heard a single agilist I consider knowing his (or her) stuff say that the goal should be 100% test automation and 0% manual testing. What I have heard people say (and what I've said myself as well, from time to time) is exactly what Jonathan is suggesting: that all regression tests should perhaps be automated but that context-driven, risk-based, exploratory manual testing should be carried out in addition to the automated tests.
Computers are good at finding bugs we told them to look out for. Humans are excellent at finding those we couldn't see coming.
That's a good point, Jonathan. "Most" would likely be a safer choice of words in many situations. Especially in those where the medium used for transmitting the message is the mostly-one-way-with-delayed-interaction kind such as weblogs or email.
Missing the slight movement of the eye-brow when you've said something can mean the difference between the recipient getting your message mostly correct or mostly incorrect.
Oh, and thanks for your own kind words as well. Much appreciated.







