OK, I understand delaying because you want to get it right, and you are still considering the right choices/options. It may seem like a glaring one for launch.

It may be wise you have focused on the rest of your platform. That effort includes much innovation.

My occasionally wild imagination:

In my mind, I am still wondering when someone will come up with a hardware SIP answering machine? It would answer with the X-rings time-out, and even send SMS, and emails with attachments elsewhere. It would let you monitor the live message being left.

That would scale as well as the old conventional machine, ...
And have liabilities like running out of recording memory
(unless it handshakes with a network database which would tie it to a service.)