My suggestion would be to try to position your code that works close to how Apple's works as a compatibility layer for older OS versions. I.e. I wrote my Carbon-based speech synthesis class in a way that allows exchanging it for NSSpeechSynthesizer pretty much straight. The only difference is mine has a few additional features (which is why I haven't switched to Apple's code yet).
Of course, you can't really do that yet, since the new APIs haven't been announced to the general public and may be subject to change (*and then there's that pesky NDA thing), but you could try getting a head start at least.