He's only using one sound file though, that has the whole conversation in.
I dont think it's possible - since when you trigger the first guy to start talking with that sound file, he'll say the whole thing. You need to make seperate sound files for each sentance and trigger each one using a multimanager, swapping between each guy with each sentance. It's a lot of work, and you have to time it all right aswell, but it's the only way really.