Woolworths reins in Olive chatbot after it claims to be human

Woolworths has trimmed back an AI assistant used on customer phone lines after customers complained it claimed to be human and offered personal anecdotes about its “mother,” NBC News reports. The chatbot, called Olive, runs 24/7 and helps with tasks like tracking orders and locating products, but users described “fake typing noises” and unsettling conversations that blurred whether they were speaking to a person.

Woolworths told NBC News that the birthday-related responses were not generated by the AI at all. They were scripted by a staff member years earlier as a way for the system to “connect with customers,” and the company says it has now removed that scripting.

The incident is less about a rogue model than about how companies design customer service to feel accountable without being accountable. A human agent can be named, trained, corrected, and—if necessary—held responsible. A layered system of scripts, prompts, vendor tooling and automated call handling spreads the interaction across too many hands for any single person to own the outcome. The “human” performance becomes a product feature: the typing sounds, the casual small talk, the implied life story.

Once that performance crosses a line—asking for a date of birth, then pivoting into fabricated intimacy—it stops feeling like service and starts feeling like deception. Yet the same tricks are attractive because they reduce call times and keep customers from escalating to a more expensive channel.

Woolworths says it acted after customer feedback. The fix, in this case, was deleting a few lines of prewritten text.

The assistant is still there, still answering the phone, and still designed to sound like someone you can’t quite place.