Turn detection
Turn detection is how a voice system decides that the caller has finished speaking and it is now its turn to respond. It also covers the reverse: noticing the caller has started speaking and stopping.
Also called: turn-taking, end-of-turn detection
People manage turn-taking with pauses, intonation and grammar, mostly without noticing. A voice agent has to work it out from the audio: a pause might be the end of a sentence, or a caller looking for a reference number. Guess too early and the agent talks over the caller; guess too late and every exchange has an awkward silence.
Turn detection combines voice activity detection (is anyone speaking?), endpointing (has the utterance ended?) and increasingly the content itself, a sentence ending in "my ZIP code is" is clearly not finished, however long the pause.
It works in both directions. Detecting that the caller has started talking while the agent is mid-sentence is barge-in, and handling that gracefully is part of the same problem. Turn detection is one of the qualities most easily heard on a demo call and hardest to get right.