Australian Retail Giants Bet on AI Shopping Assistants Amid Chatbot Mishaps
Major Australian retailers, including ASX-listed companies Woolworths, Coles, and Wesfarmers, are advancing plans for agentic shopping assistants powered by artificial intelligence. These sophisticated AI tools aim to transform customer service by handling tasks like meal planning, party organisation, and automated shopping. However, recent incidents involving primitive chatbots have raised concerns about the technology's readiness and potential for errors.
The Rise and Risks of AI in Retail
AI chatbots have become increasingly common in customer service, but their evolution from basic, rules-based systems to advanced, generative models has been fraught with challenges. Early versions, such as those described by Professor Uri Gal from the University of Sydney, relied on decision trees to provide standardised responses. In contrast, newer AI-powered bots, often built on large language models like ChatGPT, can learn and generate varied answers, offering more flexibility.
The next frontier is agentic AI, which operates autonomously to achieve objectives without constant prompts. Gal warns that this ambiguity introduces risks, including privacy issues and governance gaps, as seen in recent mishaps. For instance, Woolworths' virtual assistant, Olive, faced backlash when it attempted to relate to customers on a human level, leading to complaints about its quirky personality. The supermarket has since dialled back these features, highlighting the delicate balance companies must strike between relatability and reliability.
Notable Chatbot Failures and Legal Implications
Several high-profile cases illustrate the pitfalls of AI in retail. Bunnings, owned by Wesfarmers, faced criticism when its chatbot provided illegal electrical advice on rewiring extension cords. Similarly, Air Canada's virtual assistant incorrectly promised a bereavement fare refund, resulting in a lawsuit where the airline was held liable despite claiming the chatbot was a separate legal entity. Professor Jeannie Paterson from the University of Melbourne emphasises that companies are clearly responsible for their AI systems, and failures can lead to significant financial losses if not properly managed.
Guardian Australia's testing of retail chatbots revealed ongoing issues. Uniqlo's virtual shopping assistant struggled to recognise queries, while Woolworths' Olive gave irrelevant responses to simple questions. These examples suggest the technology remains in its infancy, with businesses implementing strict guardrails to prevent rogue behaviour, albeit at the cost of flexibility.
The Future of Agentic AI in Shopping
Despite these challenges, retailers are optimistic. Accenture's 2024 report encourages a delightfully human approach, and Woolworths has partnered with Google to enhance Olive using the Gemini model for more complex tasks. However, experts caution that without robust governance and testing, agentic AI could lead to widespread errors, such as incorrect purchases or data breaches. As companies navigate this evolving landscape, the success of AI shopping assistants will depend on balancing innovation with risk mitigation to avoid further public relations disasters.
