What happens when the AI can't answer a question?

It hands off instead of guessing: the agent takes a detailed message, flags the conversation for human follow-up, or transfers the call live — whichever you've configured for that situation. A well-built agent is explicitly trained to not improvise answers it doesn't have.

Every escalation arrives with the full transcript and the caller's details, so your team picks up mid-story instead of starting the customer over.

Escalations are also the improvement loop: recurring "couldn't answer" topics get added to the knowledge base, so the handoff rate falls month over month.