Terence Eden sent Ocado a detailed report about broken calendar invitations and received a reply he suspected was AI-generated. He complained through the satisfaction survey. Days later, a longer apology promised a real person was typing and thanked him for the diagnostic.
- Eden accepts AI for sorting messages, summarising complaints, and spotting recurring problems.
- He objects to generated reassurance standing in for someone engaging with a genuine complaint.
- He found the second reply more human, but the essay verifies neither message’s authorship nor a fix for the calendar bug.
The complaint is about whose time gets spent. A customer who has already supplied a diagnosis should be able to reach someone responsible for acting on it.
The 103-comment thread on Hacker News adds a failed support interaction and challenges Eden’s test for success.
What the thread adds
- swatcoder — on measuring reply speed: “A fast, shallow reply doesn’t save a customer’s time. It’s just a useless volley back to the customer that keeps their request off the support queue.”
- advisedwang — after describing 21 messages with an AI that failed to generate a shipping label: “Then I wrote "escalate to a human" and an hour later someone replies with a shipping label directly.” This is their account.
- tln — dissent from the blanket objection: “I think that good customer service can use AI. It’s not AI composed replies that makes the customer service bad.”
- mixedmath, PcChip, and 0c3ca83 — all question the follow-up’s authorship. As mixedmath puts it: “The second response reads as very much written by AI to me too.”
Handles are pseudonymous; HN publishes no per-comment scores. The source ordering is HN’s ranking; this selection is a slice of the thread, not a consensus.