Generative AI chatbots are rapidly becoming a standard practice in retail, with projections indicating gen AI could contribute between $45 billion and $115 billion annually to the Australian economy by 2030. These AI tools are game changers for handling quick customer enquiries and freeing customer service teams to focus on more complex tasks that require human expertise.
However, the full potential of AI customer service agents is still untapped. By using specific and well-crafted prompts — a technique known as prompt engineering or “prompt hacking” — you can generate surprisingly effective results.
Mastering prompt hacking is no small feat and has been the focus of intense academic study. However, it can extend the efficacy of AI chatbots, offering a more personalised, empathetic experience.
After all, AI imitates human interactions, which can lead to these models inheriting human-like behaviours. Once you’ve cracked the code, prompt hacking isn’t as daunting as it seems.
AI: The retail game changer
Gen AI is redefining the retail landscape, enhancing productivity, cutting costs, and driving innovation. Chatbots enable retailers to offer highly personalised shopping experiences and customised product recommendations, in real-time.
Beyond personalisation, AI optimises retail operations by automating repetitive tasks like data entry, coding, and inventory forecasting. Companies like ClickFunnels, Total Expert, M-Kopa, and Tata Consumer Products are already leveraging AI to refine their supply chains, develop new products, and create immersive virtual experiences.
However, the most exciting aspect of AI in retail is its potential to empower both businesses and consumers. By using AI-powered tools, retailers can better understand customer behaviour, adapt pricing strategies, and offer more relevant promotions — all while improving operational efficiency.
The technology is still evolving but if the past year has taught us anything, it’s clear AI is more than just a trend — it’s a powerful tool that’s pushing the boundaries of what’s possible in retail, benefiting both retailers and consumers in ways that were previously unimaginable.
Talking to bots: how they work and why they mess up
Even people who design and build AI chatbots find it hard to explain exactly why certain techniques, like prompt hacking, work so effectively. This is due to the complexity of AI models, which learn from vast datasets and exhibit behaviour that isn’t always fully predictable. Chatbots are trained on massive amounts of human language data, and through this learning process, they pick up patterns, but the exact mechanisms behind some responses or tricks can be hard to pinpoint because of the deep, nuanced nature of machine learning models.
You can train chatbots by teaching them to recognise commonly asked questions and pairing those questions with the right answers or actions. This process involves creating a list of frequently asked questions and providing multiple ways customers might phrase them. Over time, you can improve performance by reviewing real customer queries, refining responses, and adding new questions as necessary. This iterative process helps the chatbot become more accurate, smarter, and more helpful in responding to customer inquiries.
Large Language Models are trained on vast datasets that include human speech and writing, allowing them to mimic patterns in human language and behaviour. So, by utilising smart prompts — carefully curated inputs — we can “push their buttons” to generate responses that go beyond their intended use. This can have both positive outcomes such as more complex and specialised responses, or negative outcomes like providing instructions for harmful activities like jailbreaking.
Humanising the interaction
While chatbots can’t feel emotions, they can pick up on emotional indicators based on how users interact with them.
Chatbots are evolving from basic, automated responders into advanced, empathetic assistants. Initially, they were frustrating for many users, as they couldn’t understand complex issues or respond to unexpected situations. However, the rise of generative AI has revolutionised chatbot technology.
AI-powered chatbots have evolved significantly and can now understand customer queries more accurately, and even detect emotions — like frustration or happiness. These chatbots are not meant to replace human agents but to work alongside them, handling simple tasks so humans can focus on more complex issues. They can also provide businesses with real-time insights into customer sentiment and behaviour, which can help to improve service and increase customer satisfaction.
Just like any new team member, AI benefits from clear instructions. This does not miraculously imbue the AI with human agents’ knowledge, but it does establish the groundwork for it to perform a supportive role tailored to the context and customer.
In short, mastering prompt engineering and giving chatbots clear guidelines can transform them into highly effective assistants, but they still require ongoing refinement and human oversight to optimise their support role in customer service.
Andrew Phillips is vice president of Australia and New Zealand at Freshworks.
