It’s the time of year when retailers and brands most want to stand out. How can they attract customers and convert their shopping ‘wants’ into solid sales? A few years ago, the answer seemed to be using AI. After all, it can speed up processes. So why are those who adopted early AI models now scratching their heads and wondering why it hasn’t really worked to differentiate them from everyone else?
The answer is that generic AI is just that, generic. It has made ecommerce sites faster, but no different. Consumers want to shop with a brand that has distinction, that stands for something and that they can relate to. Generic AI tools used without context or industry insight, or without any clarity about what a brand means, do the opposite. In fact, they are likely to contribute to a bad shopping experience.
AI used well must first absorb and analyse the data on which it is built. In a retail context that means providing the tool with information about the brand, the sector in which it operates, and existing and potential customers. AI can only work with what it is given, leading us to problems with prompts.
The AI Prompt Challenge
It would be easy to think that crafting a prompt for an AI tool such as ChatGPT is about asking a question – in much the same way that we ask questions of Google. But this won’t deliver the results that are needed because AI tools need specific commands.
During the development process, prompts need to be engineered with care because they have the potential to deliver not only faster responses to explicit prompts but for business users, a significantly more rewarding outcome. That’s why the engineering of prompts is top of mind for retailers and brands that are looking to create faster, more memorable experiences.
In the meantime, trying to manage prompts with generic AI is proving time-consuming, requiring business users to complete prompt forms with a multitude of words and multiple fields.
Getting prompts right
Fortunately, AI products are being developed that move away from the generic and aim to deliver intelligent – and genuinely helpful – prompting tools. These will be solutions that don’t require ecommerce merchants and their teams to have expertise in prompt engineering, but will open the door to AI success anyway.
This type of integrated, instinctive prompting tool will provide the essential context that AI requires to deliver quality content that can be used on ecommerce sites, without making the process complicated. It will ultimately represent the difference between AI that is fast, and AI that is intuitive and relevant…and fast.
All marketers are familiar with writing a brief that aims to create impactful and persuasive content. Most now do this manually and then give it to someone in the creative team to develop until all parties are happy with the result. This is a process that can take time and a great deal of thought. Using an intelligent engine to prompt AI, however, augments what the marketing and creative experts do now, with contextual relevance, but in a fraction of the time.
In summary
What retailers need to think about now is not how they can improve their prompt expertise to maximise their existing AI tools, but instead how much easier it would be to specify and deploy an AI model that can quickly master every aspect of their brand, their business and the broader sector. These models will emerge in 2024 to deliver AI that can be harnessed by retailers and integrated fully into their processes to deliver better experiences for customers. This will give them more control, allowing them to achieve the essential differentiation they need to stand out from the crowd.
Contributor: By John Williams, CTO, Amplience
