The world’s largest search companies — so far, Microsoft, Google and Baidu — are the pioneers of generative AI. However, one internet giant that has not yet entered the conversation is Amazon, which stands to benefit tremendously from generative AI-powered search.
Amazon: the Ecommerce Gorilla Still Behind in Generative AI
Why is generative AI crucial for Amazon? Amazon is the 13th-most visited website worldwide, making it the world’s largest search engine focused on shopping. Although overlooked, search is one of Amazon’s most critical technologies because the vast majority of Amazon purchases originate from Amazon’s search bar, so Amazon has to get search right. Indeed, Amazon takes search seriously and has an entire subsidiary called A9, referring to the nine letters in the word ‘algorithm,’ focused on building leading ecommerce search technology.
The Transformative Potential of Generative AI in Ecommerce search
Generative AI technologies like ChatGPT will have a seismic impact on ecommerce search, because it solves the critical problem of how to accurately match user search intent with the right product results.
Presently, ecommerce search is primarily keyword-based. For example, if I search “Which TV should I get?” on Amazon, the top three results are a Dr. Seuss book called “What Pet Should I get?,” a Sony TV and a TV show “I Should Have Known.” The search bar can’t understand my intent because it is simply matching keywords.
ChatGPT will fundamentally change this approach to search, because its natural language processing system can understand user intent, thanks to an unprecedented predictive model trained on nearly a trillion words.
Here’s an example: When I type “What TV should I get?” in ChatGPT, I get a detailed five-bullet-point response with top considerations for choosing a television, such as budget, size and features. It also suggests several leading brands, such as Samsung, Sony and LG. ChatGPT understands what I’m trying to find.
However, ChatGPT’s weakness is that it is limited by the data available to train its language model. Because ChatGPT uses publicly available information, it can only return a generic set of suggestions for choosing a suitable TV. There is no product-level personalization based on data such as pricing, popularity and users’ purchase history. ChatGPT needs access to proprietary platform data from retailers in order to make product-level personalized recommendations.
Therefore, the next big step in ecommerce search will be to combine the intent prediction capability of generative AI with the massive troves of user data on ecommerce websites. Only then can ecommerce companies develop a next-generation search experience that can accurately return search results based on both user intent and external data. The ultimate goal of ecommerce search is surfacing the right product for customers without them ever having to use a third-party search engine at any point, and generative AI is a big step toward this goal.
In practice, here’s what next-generation ecommerce search will look like:
First, users can use natural language in the ecommerce search bar. For example, currently Amazon does not understand the query, “What books are in the Harry Potter series?,” which returns irrelevant top results such as a LEGO Marvel Spider-Man toy and a Star Wars encyclopedia. With generative AI, the search results will list the individual Harry Potter books in order of their sequence, and the user never has to search for this information on Google or Wikipedia.
Second, search will deliver next-generation product recommendations. Generative AI-enabled ecommerce search will integrate both on-platform data, such as product and user data, as well as off-platform data, such as social media signals and product reviews from influencers. Currently, ecommerce recommendations rely almost exclusively on the retailer’s own data. With generative AI, retailers can access an almost limitless trove of external data, which will combine with on-platform data to deliver a significantly more relevant and personalized set of product recommendations.
Third, chatbot functions will integrate seamlessly into the search bar. Today, an ecommerce website’s search bar and chatbot are separate elements, because product search uses keyword matching while the chatbot responds to a rudimentary set of natural language queries. This separation creates a clunky UX in which chatbots are often located on the bottom corner of the screen. However, with the advent of generative AI, the search bar and chatbot can respond to the same set of natural language queries. This allows the chatbot to be unified seamlessly into the search bar, which can both find products and answer general, non-product related questions.
Ecommerce is the Next Frontier of Generative AI
Following the world’s biggest search engines, the next major wave of generative AI adopters will be the world’s leading ecommerce companies, which will combine generative AI’s ability to understand user intent with their proprietary platform data to turbocharge ecommerce search capabilities. This will create a fascinating new dimension in the generative AI race.
Bob Ma is an Investment Manager at WIND Ventures, where he invests in retail, fintech, consumer and energy technologies.