Because the context is dropped. The customer explains the problem to the bot, gets handed to a person, and is asked to explain it again from the beginning.
By then they have spent five minutes and are less patient than when they arrived, and the agent is starting from nothing.
Handover passes the full conversation, what the assistant attempted, which sources it used and why it stopped, into a ticket in Desk on that customer's record.
Triggers are rules rather than judgement: an unciteable answer, a blocked topic, an explicit request, or a customer tier that always goes to a person.
The agent opens with what has been tried and what the customer actually wants, so the conversation starts where it left off.
It also produces evidence: the questions that consistently trigger handover are the articles you have not written.
It will not hold a customer in an AI conversation. Asking for a person is always honoured immediately, with no deflection loop and no requirement to explain why.
Outside staffed hours it takes the ticket and says when somebody will reply, rather than pretending help is available.
Nothing here is an integration. Each line is two products reading the same record from different sides.
Handover passes the full conversation, what the assistant attempted, which sources it used and why it stopped, into a ticket in Desk on that customer's record.
The agent opens with what has been tried and what the customer actually wants, so the conversation starts where it left off.
It will not hold a customer in an AI conversation. Asking for a person is always honoured immediately, with no deflection loop and no requirement to explain why.
Fourteen days, every module, no card. Or half an hour with someone who will run it on your own records and tell you where it does not help.