What good actually sounds like on a phone call

People judge a phone agent in about four seconds, and almost none of that judgement is about the voice. It is about timing.

Silence is the enemy

When an agent looks something up — checking availability, say — it has stopped speaking and cannot speak again until the lookup returns. To the caller that is dead air, and dead air on a phone means the line has dropped. So they repeat themselves. And a naive system treats the repeat as a second request, and books the table twice.

The fix is layered. The agent says something before it goes quiet. Identical requests still in flight are folded into one. And an action that lands twice is recognised as the same action rather than performed again.

It has to be interruptible

Real conversations overlap. A caller who has heard enough starts talking over the greeting, and an agent that keeps reciting through it sounds like a machine no matter how good the voice is. Barge-in is not a refinement; it is the difference between a conversation and an announcement.

And it has to know when to stop

The most important sentence a phone agent says is the one where it admits it cannot help and puts a person on. Getting that moment right buys far more trust than any amount of polish on the ninety per cent it handles alone.