We already have them. I'm agender and I'm typing this from my workplace (during lunch).

The response to this AI voice demonstrates that a lot of people automatically hear any human(ish) voice as either female or male, one or the other. Similarly, people automatically recognise individuals as either female or male, one or the other. There's a very narrow band of presentation that won't be read automatically as one of those two options.

See how difficult it is to stop your brain from making automatic (possibly false) assumptions.