What is the belief, and where does it come from?
The belief is that automating customer support makes customers angry, and that a store which values its customers should keep people on every ticket. It comes from experience that almost everyone shares: a chatbot that asks which topic you need help with when you have already typed your question, a scripted reply that says your order is on its way when the tracking has not moved in a week, a loop of menus with no path to a person, and a survey afterwards asking how it went.
The belief is a fair reading of that evidence, and any honest examination has to start by agreeing with the experience behind it.
What is true in the belief?
What is true is that a reply which does not know the customer's order makes them angrier than no reply, because it proves that nobody looked. It is true that a system which loops a customer through the same options is worse than a queue, because a queue at least ends. And it is true that hiding the path to a person, whether behind menus or behind a cheerful bot, is experienced as contempt. These three causes explain the anger, and a scripted human agent who cannot see the order produces the same anger as the bot.
It is also true that some conversations need a person regardless of how good the automated reply is: a bereaved customer, a safety complaint, a threat. Automation that does not recognize those is a failure.
What does answering from the order change?
The Order Support FTE on ftes.ai reads the order, the shipment's carrier scans and the policy before it writes a reply, and the reply states those facts: the last scan with its time and place, the delivery estimate, the return window as it applies to this order, the exchange the policy allows. A customer who receives that reply has been looked at. The first cause of the anger, a reply that does not know the order, is gone by construction, because a reply the FTE cannot ground in the records is not sent; the ticket is escalated instead.
The reply is also written in the store's voice from the remit, without the phrases the owner banned, and it never contains a promise the policy does not make.
How is looping and hiding the human avoided?
Looping is avoided because the Order Support FTE works inside the help desk on the ticket the customer already wrote; there is no menu and no re-asking of the question. Hiding the human is avoided because escalation is a mechanism, not a last resort: anything above a limit, outside policy, on the always-escalate list, or that the customer asks a person for goes to a named person with the thread, the order, the scan, the policy excerpt and a suggested resolution attached. The customer is told a person is reviewing it and, because the person has everything in front of them, the wait is short.
Escalation accuracy, the share of escalations a person agrees were correct, is reviewed monthly, so a pattern of not escalating when it should is visible and corrected.
Should customers be told they are talking to an FTE?
Whether customers are told is the store's decision, set in the workspace, and both choices are used. Some stores have the Order Support FTE draft every reply for a person to send under their own name, which is the Level 0 default anyway. Others let it answer routine order-status questions directly and sign as the store's support team, with a person for everything else. The honest position is that the choice belongs to the store and that the reply's quality, not its signature, is what customers react to.
What no store should do is pretend a scripted reply is a person. That is the practice that earned the belief.
How can a store check for itself?
A store can check the belief on its own tickets with one question: does each reply state a fact from the order that the customer could not see themselves? Read ten replies, from a bot, from an outsourced agent, from an in-house agent, from an Order Support FTE, and count. Then check two more things: did any reply send the customer back to a menu or a form they had already used, and could a customer who asked for a person get one within the reply itself? Replies that pass all three do not make customers angry, whoever wrote them.
What remains true is that some tickets need a person from the first word, and that a store which does not staff those, in-house or outsourced, will make customers angry with or without automation.