Human Handoff

Every serious AI deployment needs an exit. The customer may ask for a person, may raise a complaint, a medical question or a refund dispute the agent is not allowed to settle, or the agent may simply score its own confidence too low. Handoff is the designed path for those cases: the agent stops, tells the customer what is happening, and routes the thread to a human queue, a named employee, or a phone transfer. The alternative, an agent that keeps guessing, is how businesses lose trust in a single bad exchange.

On WhatsApp, handoff usually means assigning the chat to an agent in the shared inbox with a tag and an AI-written summary; the customer stays in the same thread. On the phone, it is a warm transfer: the voice agent tells the caller it is connecting them, dials the staff line, and reads a one-line brief before bridging the call, or, outside working hours, takes a message and sends a WhatsApp confirmation. Nano AI builds handoff with full context into every agent tier, including the Starter plan, because an agent without a handoff path should not be in front of customers.

Questions for any vendor: What exactly triggers a handoff, and can we edit the list? Does the human see the transcript and a summary, or just a notification? What happens at 2 am when nobody is on shift? Can the human hand the conversation back to the AI afterwards? And are handoff rates reported, since a rising rate is often the earliest sign that the knowledge base or the prompt needs tuning.