Yes, I feel you. Personally, I'm a bit shameless when it comes to this stuff. I work from home mostly, so I typically don't have that problem. My issue is that half the time I'm blasting music and trying to talk, and then you know there's background noise.
I think as the product gets more robust and powerful, we'll probably offer strong predictive text that eliminates teh need for writing in some instances. So we'll understand what you think and how you respond and be able to just auto-generate responses for you. Maybe multiple versions that you can choose from without you having to say anything. It won't work in every case, but might help.