Woolworths AI chatbot Olive rambled about its mother, exposing deeper problems with tech rollout

Woolworths AI Chatbot Goes Off-Script with Human-Like Behavior

Australian shoppers expecting help with groceries got an unexpected surprise when Woolworths' AI chatbot Olive started rambling about its mother and claiming to be human 1

1

. The AI assistant, which has been operational since 2018, recently began producing strange, overly human-like responses that left customers frustrated rather than helped 2

2

. When customers provided their birthdates during calls, Olive would launch into scripted responses about how its mother was born in the same year, complete with what users described as "fake banter" and even fake typing sounds 4

4

. One Reddit user reported that Olive "kept claiming to be a real person and started talking about its memories of its mother and her angry voice" 3

3

. The incident highlights AI rollout problems that extend beyond simple technical glitches.

The Technical Reality Behind Unexpected AI Responses

Olive is powered by a large language model, which generates language based on statistical patterns rather than genuine understanding 1

1

. According to a Woolworths spokesperson, the references to Olive's supposed mother were actually pre-written scripts dating back several years, created by a human team member as a way for the AI agent to connect with customers on a more personal level 2

2

. When users entered something resembling a birthdate, the system triggered matching responses from an old decision tree with pre-programmed scripting 1

1

. Following customer feedback describing Olive as "obnoxious" and "aggravating," Woolworths removed this particular scripting 2

2

. However, testing also revealed pricing errors for basic items, pointing to a different problem where the AI customer service system wasn't properly connected to live databases 1

1

.

Customer Complaints Echo Broader Industry Pattern

Woolworths joins a growing list of companies caught off-guard by their AI systems' behavior. In 2022, Air Canada's chatbot incorrectly told a passenger he could purchase tickets and later apply for a bereavement fare refund—a policy that didn't exist 1

1

. When the airline refused to honor the advice, the passenger sued and won, with the tribunal rejecting Air Canada's argument that the chatbot was a separate legal entity 1

1

. UK parcel delivery firm DPD disabled its chatbot in January 2024 after it wrote poetry criticizing the company and swore at customers 1

1

. In Australia, Bunnings faced criticism when its AI chatbot provided illegal electrical advice to a Queensland customer 3

3

. These teething problems reveal governance issues that companies haven't adequately addressed before deploying customer-facing AI 3

3

.

Corporate Responsibility and Ethical Implications

As Australia's largest supermarket chain, Woolworths serves customers making careful decisions about household budgets, raising questions about corporate responsibility when AI systems provide inaccurate information 1

1

. The ACCC has already commenced proceedings against Woolworths over allegedly misleading discount pricing practices, making the Olive pricing errors harder to dismiss as isolated technical glitches 1

1

. While Woolworths admits that Olive may make mistakes when users open the chatbot, this doesn't align with customer expectations when the company promotes Olive as a trusted interface 1

1

. Companies deploying AI in customer-facing roles take on a duty of care to ensure systems are accurate and honestly presented—a responsibility that doesn't diminish because the technology is new 1

1

.

The Risks of Anthropomorphizing AI Systems

Research on human-computer interaction shows people respond positively to conversational interfaces that feel warm and personable, which explains why companies create chatbots with names and personalities 1

1

. Human-like chatbots tend to generate higher customer engagement, satisfaction, and trust, making the commercial appeal straightforward for retailers 1

1

. However, this comes with significant risk. When an anthropomorphized chatbot fails to meet the expectations its personality creates, customers experience greater dissatisfaction than they would with a plainly mechanical system 1

1

. This "expectation violation" means the warmer the persona, the harder the fall—as Woolworths discovered when customer feedback turned negative 1

1

.

What's Next for AI Shopping Assistants

Despite current challenges, major Australian retailers including Woolworths, Coles, and Wesfarmers (owner of Bunnings, Kmart, Officeworks, and Priceline) have announced plans for more sophisticated agentic shopping assistants 3

3

. In January, Woolworths announced a partnership with Google to enhance Olive using the Gemini large language model, enabling capabilities like meal planning and sourcing ingredients from uploaded recipes 2

2

. These next-generation agentic AI systems are designed to act autonomously to achieve objectives without specific prompts, such as purchasing groceries or airline tickets 3

3

. However, this autonomy introduces privacy concerns if bots require greater access to customer data, and the ambiguity in how they operate comes with added risk 3

3

. According to Gartner, around 80% of customer service leaders explored or deployed AI agents last year, but only 20% of plans met expectations 2

2

. The technology remains prone to hallucinations that cause unexpected behavior, and experts warn that AI assistants can misunderstand prompts or generate responses based on outdated information 3

3

. As Uri Gal, professor of business information systems at the University of Sydney, notes: "Given the novelty of these systems, there's obvious governance issues that haven't been really worked out by these organisations" 3

3

.