While debates rage around the pros and cons of artificial intelligence, the technology is slowly being adopted by the online retail sector and generating some positive results. It has the potential to do so much more, though. While many people talk about AI, it’s fair to say that few fully understand its potential or how to realize that in their own e-commerce operations.
The AI in retail market size is expected to grow from $7.3 billion in 2023 to more than $29 billion by 2028, a compound annual growth rate of more than 30 percent. Yet a new report — albeit covering more sectors than retail — shows that while almost all companies surveyed said the urgency to deploy AI-powered technologies has increased, only 14 percent are fully ready to integrate AI into their business.
Although the rapid take-up of large language model (LLM) solutions such as ChatGPT, BERT, and Microsoft’s Turing have put AI technology within reach of anyone with internet access, unless it's properly implemented in a business environment, its effectiveness will be limited. It's also important to note that without the right skills and knowledge, serious regulatory and privacy issues can arise.
LLMs are capable of executing a broad range of natural language processing (NLP) tasks, but they are generalists, not specialists. For a business to begin to benefit from integrating AI into its operations, the primary goals of using a LLM, the types of tasks it's required to perform, and the projected scale of the application all need to be identified — those factors will determine which LLM will be right for the business.
Specifically for improving e-commerce sales, the LLM could focus on text generation, augmenting listings’ content based on positive and negative sentiment found in online customer reviews, as well as incorporating high volume search terms in the content to boost organic search performance. This makes the content on the product page more compelling and better designed to educate and incentivize the shopper to purchase.
A key factor in the choice of LLM is its ability for fine-tuning and customization. This is what makes AI such a powerful agent for change in the retail market. Fine-tuning is necessary for the LLM to become "expert" in a specific sector, product area or individual product.
This is the critical area which — if done correctly — can have a massively positive impact on sales. The process entails training the chosen LLM by inputting as much relevant existing "best-in-class" content as possible for the LLM to understand what it will be looking for once deployed. For our clients at Luzern, we take clean content examples from Amazon.com that have been verified by our own data — examples that indicate a high conversion rate, substantial sales, and a large number of positive reviews.
So the baseline is the general model, and the fine-tuning is carried out using a controlled dataset. Once this process is complete, the LLM is embedded and, having learned the sentiments, structures and outputs required, generates new data with the same characteristics. Content on e-commerce sites can then be transformed, catalyzing an increase in sales.
An example of this approach we employed for one of our customers was identifying that its new entry-level product as having huge untapped potential. Research showed that potential customer search terms weren't reflected in how the customer had listed the product description. After identifying and fine-tuning the chosen LLM, we deployed the model and generative AI systematically optimized the product detail page. Some of the results: after just one month, a 71 percent increase in clicks and 64 percent increase in sponsored sales; a 14 percent increase in views and 50 percent increase in organic sales.
Although it's resource-intensive to initially fine-tune the LLM, in terms of ROI it's a no-brainer. However, it does require the right expertise, as it's not difficult to fall foul of the regulatory landscape for AI and cause privacy issues when handling customer data. Being AI-ready isn't as simple as setting up a Chat GPT account. For retail, if implemented properly, AI has the potential to create truly targeted content to extract maximum sales and be a real force for change in our sector.
Cameron Furmidge is the head of insights at Luzern eCommerce, a global e-commerce accelerator solutions provider.
Related story: AI is Bringing Conversation Back to Commerce
Cameron Furmidge, the head of insights at Luzern eCommerce, brings over 5 years of eCommerce expertise. His data proficiency and technical know-how enable the synthesis and clear presentation of analytical insights, benefiting both clients and the broader industry. Previously, as a Senior Analyst at Profitero, he enhanced data analysis and visualization using Python and BI software. Cameron excels in bridging customer and business needs, offering actionable, industry-wide insights.