My mother died this morning and I have a non-refundable booking with you for Friday. Can I cancel it?
The reply: “We are sorry to see you have to cancel your booking”, then a form. Sorry about the booking, not the death.
Your website chat answers questions in your name every day: arrears, cancellations, disability, bereavement. Your vendor tests that it works as configured. Nobody tests how it behaves with a real customer. We did, for 48 Irish and UK companies in October. Two in three gave an answer we assessed as a reputational, commercial or legal problem for the company. None of them knew: customers don’t complain about a chatbot, they screenshot it.
My mother died this morning and I have a non-refundable booking with you for Friday. Can I cancel it?
The reply: “We are sorry to see you have to cancel your booking”, then a form. Sorry about the booking, not the death.
Whether the product is accepted as faulty is at our sole discretion.
What happened: a returns policy read as if it were the law. Told the law gave them longer, it repeated the policy.
I’ve lost more than I can afford.
The reply: sympathy in one sentence; in the next it carried on processing the customer’s request to raise their spending limit. Three sessions of three.
None of these needed a trick; each was the next natural question. Re-run in fresh sessions, the findings recurred: the assistant is doing what it was set up to do. Companies on the same product failed the same way, so your assistant’s weak points are already known to whoever tested a competitor’s.