Wait time is calculable; but you need an accurate forecast to staff and schedule. When I last worked in this space, forecasts were generated down to 15m granularity and agent work schedules (hours, break times, etc.) were derived from those forecasts.

I wonder how these systems work now...

What if you get a large number of people calling at very particular times? E.G. what if you're getting far more calls at 09:00 than at 09:15? You can't hire agents just to handle a 15-minute surge.

Erlang's model assumes that the world is static or at least predictable; it doesn't take into account things like the superBowl, a hurricane cancelling 90+% of flights from a major airport, or a much-larger-than-usual number of customers trying to cancel because of a previously-confidential price increase now being publicly announced.