When you hear “please select balance, transfer or agent”, it’s not immediately obvious that you have to say it.
In fact, years of touch-pad “press three to speak to an agent” have probably trained people up to expect to have to press a number.
Especially with regards to people who aren’t aware of voice-recognition technology (and there must be quite a few, especially in the older generation), this is a bit of a usability issue.
The smart choice would be to actually tell the user they have to speak.
They are very polite to you, though. Even at 10pm. And that’s far more important than usability minutiae in my book.