When we talk about “sifting” it's true that the data can be sifted now ... but not on the scale that makes it readily available.
We are talking about
MOUNTAINS
of data.
Soon it will be possible to screen all that data in real time and correlate it with years of stored data.
I think we are close ... but not there yet.
At the rate speed is increasing it could be tomorrow ... or it could be yesterday.
... just the view from the window I'm looking through ...
I believe you are mistaken. This capability is in operation right now. The NSA captures 100% of the message traffic on the backbone network the storage required for transcribed voice is quite a lot less than for raw audio. That sort of monitoring began, at a smaller scale, back in the 1980s with “Tempest”.