The difference between a transfer and an abandonment is about thirty seconds of context. Warm transfers — where the AI agent briefs the human before connecting the caller — shipped in v4.1, and the data since has changed how we think about handoffs entirely.
The cold transfer problem
A classic cold handoff forces the caller to start over: re-explain who they are, repeat the account number, retell the problem. Callers rate these calls poorly even when the human resolves the issue quickly, because the phone system just demonstrated that it wasted their first three minutes.
How warm transfers work
When an AI agent decides to escalate, it keeps the caller on a brief hold while it whispers a summary to the receiving human: caller name, account, what they need, what has already been tried, and the agent’s read on urgency and sentiment. The human accepts, the bridge completes, and the full transcript and notes attach to the call record automatically.
The whisper is generated from the live conversation state, not a template — a billing dispute brief sounds different from a panicked outage call, because it should.
What the numbers say
Across early adopters, warm transfers cut post-handoff call duration by 38% and raised transfer-call CSAT to parity with calls humans answered directly. The surprising second-order effect: human agents report trusting the AI more, because every handoff arrives with proof the agent did real work first.