OSX has two different things you might be talking about built in:
There's MacInTalk, the speech-generator that can be used to read text outloud for various reasons, which has been part of the MacOS since the early 90s at least (and sadly, hasn't had the voices updated since maybe the mid-90s, but it's still quite understandable).
Then there's the speech recognition engine, which has also been built-in to the MacOS for quite some time (since well before OSX), but isn't as powerful as what you might be picturing. It is NOT a text input device, like some software you can buy--it is a simple recognition engine designed so you can activate commands or scripts by talking to your Mac. I haven't used it in years because it tended to trigger unintentionally when you were having a conversation near your Mac, but it worked ok if memory serves.
If you want to talk instead of typing, you're going to have to buy a piece of software to do that. MacSpeech sells one option, but there are others I believe.
Automator is a tool designed for people who want to automate a repetitive task but know nothing about scripting or don't want to bother with it. It's incredibly powerful, but only if you actually have need for it--I, for example, have only done something that it would have been useful for once in the past couple of years, and it was sadly before 10.4 existed. Others (particularly people who process a lot of files, like photographers) could get immense benefit from it--just lay out a set of actions and let it take care of the rest. Apply the saved actions to a folder, and it'll process whatever you drop in there.
Incidentally, I've become rather familiar with the speech generation system on the Mac recently since I've been largely unable to speak for the entire year to date, so I've been using it to pick up some of the slack instead of just writing notes or typing. My wife keeps telling me--or "Bruce", the voice, rather--that I have a funny accent.
