Automationabout 15 min

    How do I hand a conversation from an AI agent to a person?

    Short answer

    Define the triggers that end the agent's turn: frustration, a request for a human, an out-of-scope question, or repeated confusion. On handoff, notify the right person, hand over the full transcript, and stop the agent so the customer is never talking to both.

    Before you start

    • A running agent handling conversations.
    • Who takes escalations during hours.
    • What happens after hours.
    • Your response time promise.

    Step by step

    1. 1

      List the escalation triggers

      Asking for a human, frustration, complaints, out-of-scope requests, and two failed attempts at the same question.

      Vague escalation rules mean the agent keeps trying long past the point of usefulness.

    2. 2

      Always honor a direct request

      If someone asks for a person, the agent stops immediately and says a person is coming.

      Refusing that request is the fastest way to lose a customer.

    3. 3

      Pass the whole transcript

      The person picking up sees everything already said.

      Making a customer repeat themselves wastes the whole interaction.

    4. 4

      Route by hours

      During hours, notify and transfer live. After hours, capture the request and promise a specific callback window.

      An honest wait beats a fake availability.

    5. 5

      Stop the agent on handoff

      Once a human is engaged, the agent goes quiet until the thread is closed.

      Two responders in one thread is worse than either alone.

    6. 6

      Track the handoff rate

      Measure how often and why escalation happens, and use it to improve the agent.

      Escalation reasons are a free list of what to teach next.

    What good looks like

    • Customers reach a person when they need one.
    • Nobody repeats themselves after a handoff.
    • After-hours escalations get honest expectations.
    • Handoff data drives agent improvements.

    Common mistakes

    The ways this goes wrong in automation, and what to do instead.

    No way to reach a human

    A customer trapped in a loop with a bot leaves and posts about it.

    Do this instead: Make a request for a person an immediate, unconditional escalation.

    Handing off without context

    The person starts from zero and the customer repeats the whole story.

    Do this instead: Transfer the full transcript and a short summary with the notification.

    Agent keeps replying after handoff

    The thread becomes confusing and the customer trusts neither party.

    Do this instead: Pause the agent automatically as soon as a human replies.

    Frequently asked

    5 questions about hand off from ai to a person

    For intake, a minority. Rising escalation usually means a knowledge gap you can close.

    Pick the next thing to set up, or back up a step if something here assumed work you have not done yet.

    People usually get here from

    Guides that lead into this one, in case you skipped a step.

    Was this guide helpful?

    Tell us what worked and what didn't. It shapes what we rewrite next.

    You didn't get into business to be in the tech business.

    Managed Tech clients hand this to us. We build it, test it, maintain it, and add to it as the business changes, so nobody on your team has to become the person who knows how the system works.

    Schedule a fit callCallCalls are answered by Troy, our AI assistant. Troy answers questions about the system and services and can schedule, reschedule, or cancel appointments. He never sells or makes offers. Any communication with TactStack is stored in our systems and referenced by Troy during certain interactions. AI can make mistakes.Full documentation