The pandemic permanently changed the face of e-commerce. Unable to shop in person, consumers increasingly relied upon virtual shopping experiences for their needs. But as society returned to normal, the popularity of online shopping did not decrease. This was due to rapid improvements in shopping experiences, especially search capabilities powered by generative artificial intelligence (GenAI). These improvements allowed customers to ask open-ended questions rather than needing to create specific queries for each item. Now, organizations must determine the best way to leverage GenAI and large language models (LLMs) by accounting for numerous considerations: Customer problems and the return on investment (ROI) through GenAI-powered solutions, cost and performance, legal and PR implications, and privacy.
Dramatic Growth in E-Commerce
GenAI and machine learning (ML) are nothing new in the world of e-commerce, but the pandemic turbocharged the adoption of major strategic initiatives around e-commerce in the retail industry. In 2020 alone, online retail sales grew at more than twice the rate of 2019, and the trend continues, with worldwide retail e-commerce sales expected to reach $6.3 trillion in 2024.
As a result, many companies began to emerge with GenAI solutions specifically targeted to the e-commerce marketplace. Suddenly, an area once dominated by ML saw a gradual but decided growth in deep learning, natural language processing and LLMs.
GenAI’s Impact on Consumer Interactions
Twenty years ago, online e-commerce searches meant formulating careful, thorough queries to look for a very specific product. Part of GenAI’s impact is the ability for people to express their desired intent in numerous modes (text, voice, image) while asking more open-ended questions. On leading retail sites, a question such as “What do I need for a camping trip?” will take the consumer to a page with purchase options from tents and sleeping bags to ancillary supplies, including hammocks, backpacks, canteens and even books to learn more about camping. Consumers can begin and end their search journeys on a single site.
Important Considerations for Leveraging GenAI
Ultimately, GenAI is a solution, a virtual tool used to address issues. Is GenAI the proper solution for each specific issue? The answer is: It depends. It’s crucial for organizations to understand their needs to determine how or even if they should leverage GenAI LLMs. Why use a sledgehammer when a screwdriver will do the job? Aspects to consider include:
- Cost and performance implications. It’s not cheap to run models in real time, and runtime or storage inferencing costs accumulate quickly. Additionally, optimizing performance is critical with complex (one billion-plus parameter) models. Customers are shopping for convenience and may not be willing to wait several seconds to receive search results. While costs can be initially overwhelming to organizations, it’s important that they are not limiting factors, because these costs tend to decrease over time with increased availability and familiarity.
- Legal and PR implications. Contracts and intellectual property concerns are just one consideration. If people search for inappropriate, culturally insensitive or profane terminology on a website without properly adjusted guardrails, companies may be at the mercy of the LLM, risking unwanted attention or legal action for undesirable, non-inclusive or discriminatory results.
- Readiness of enterprises to best leverage technology. Enterprises are innovating using distilled models to inform small language models (SLMs). Still wary of LLM hallucinations, these companies use enterprise-specific data to train and produce optimal output from their SLMs.
- Organizations face numerous challenges when utilizing personal, confidential data in GenAI systems. The implementation of specific mitigation strategies is imperative.
Real-World Examples and Use Cases
GenAI continues to improve the customer shopping journey through numerous innovations. Language learning leads to improved search results recognizing different words with similar meanings (e.g., pants, khakis, trousers), providing more comprehensive results for the user. Additionally, one large U.S. retailer offers conversational experiences within the search function where specific commands or requests such as “help me plan a Super Bowl watch party” yield results for everything from serving trays to snacks, tables and even large-screen televisions.
Customers’ increasing desire for efficiency and white-glove service led to the creation of shopping assistant tools such as Amazon’s new chatbot, Rufus, which became widely available in July 2024. Rufus provides customers with the e-commerce equivalent of an expert personal assistant to aid in their shopping journeys.
Sephora’s revolutionary virtual try-on lets customers test over 1,000 variations of makeup through its app, assisting each user in finding a complement to a particular outfit or a match with other accessories. Similar options are available through Nike for virtually accessing sneakers. Highlighting these products and their capabilities enhances the online shopping experience for users, while the initial attraction of trying out a brand-new technological feature ensures traffic growth.
The power of GenAI is undeniable, but it can be a double-edged sword. Companies should not expect optimum outcomes on the very first day; instead, it’s vital for them to appropriately leverage GenAI in the best interest of their specific customer base. GenAI is most effective when used judiciously, and it’s critical for organizations to recognize the potential costs—financial and reputational—of any improper usage.
In the world of e-commerce, top organizations are leading the way in GenAI utilization by investing in efficient, customized platforms. The potential for additional innovative features is limitless, providing consumers access to experiences exclusive to online shopping.
