Terence Eden, a developer and blogger, filed a detailed technical bug report with British online supermarket Ocado about malformed calendar invites. He included a clear problem statement and diagnostic analysis. What he got back was a textbook AI-generated response — sycophantic opener, em-dash flourishes, and zero evidence anyone had read his message. The tells were specific and identifiable. The response opened with performative agreement ('You're entirely right'), deployed an em-dash almost never seen in British customer service prose, and offered the generic assurance that feedback had been 'passed on.' No questions, no engagement with the technical content, no indication a human had triaged the issue. Eden, who had encountered AI-generated responses before, clocked it immediately. His countermove was deliberate: he gave 1-star scores across every metric on the subsequent satisfaction survey and explicitly called out the AI response as disrespectful. Customer satisfaction surveys are typically gamed by customers who inflate scores to protect human agents from punitive metrics. Eden broke that social contract because there was no human to protect. The result was instructive. A few days later, a second response arrived — rambling sentences, imperfect grammar, direct engagement with the bug report's contents, and an explicit acknowledgment that the first response had been robotic. The human agent thanked him for the diagnostic work and confirmed it had been escalated. The contrast between the two messages was the entire argument. Eden's broader point is precise: AI triage of inbound messages, summarization for agents, and pattern detection across complaint volumes are all defensible uses. Replying to a customer with AI-generated text is a different act entirely. It tells the customer that the company values its own time more than the customer's. The asymmetry is the insult — you spent minutes composing a detailed report; they spent zero seconds reading it. This connects to Professor Kate Devlin's AI Resist List, which Eden encountered at EMFCamp. The project's premise is simple: loudly identifying and objecting to bad AI deployments creates social pressure that quieter acceptance does not. Eden's post is a worked example of the method — complain specifically, complain loudly, and document the result. The underlying dynamic is not about AI capability. It is about corporate allocation of respect. When a company routes genuine customer effort through a language model that generates a non-response, it has made a resource decision: the customer's problem is not worth a human's time. The survey mechanism, ironically, was the only channel where Eden's objection registered as a signal strong enough to trigger human engagement.
