AI and Generative AI: numerous use cases already in place
Artificial Intelligence is more than just a trend, it's a technological revolution that has been part of our daily lives for several years now. It has become more visible to the general public with the arrival of ChatGPT, MidJourney and other models of Gen AI (Generative AI). But it should not overshadow the other types of AI that have already made significant inroads into systems, improving the user experience and the efficiency of operations.
Generative AI for e-commerce
With Generative AI, e-retailers are seeing the transformation of many of the tasks at the heart of their marketing activities: generating SEO-optimized item descriptions, translating them for different international sites, creating images and videos to feed product sheets and promotions, optimizing chatbots with enhanced conversational interactions, and creating emails with personalized content.
AI for product discovery
E-tailers have not waited for the arrival of Generative AI to apply AI to their e-commerce platforms. Machine learning, in particular, has been providing them with concrete applications for several years now, allowing them to automate features they've been secretly dreaming about. Technically, machine learning algorithms have the ability to learn from data and extract relevant information by applying statistical rules that define a learning model.
With the growth of data, machine learning models have proven their relevance in customer segmentation, which in turn is enabled for various marketing tools used on e-commerce sites to provide a better customer experience and optimize conversion. With the latest developments in AI, this personalization goes even further by analyzing data in real time to offer ultra-personalization in real time.
Many AI-powered devices are part of product discovery :
- The semantic search engine : This uses NLP (Natural Language Processing) technologies or, more recently, LLMs (Large Language Models) to interpret queries and provide more relevant results, even if the sentence is formulated in natural language. This ability to understand is particularly important because users tend to formulate their queries in real sentences rather than in an association of keywords. This evolution is due to the use of voice assistants and is also reflected in voice search (example: "I'm looking for a long dress for my friend Nathalie's wedding").
- Personalized recommendation : these systems use collaborative filtering and content-based filtering algorithms to suggest relevant products to each user, in particular by analyzing the customer's browsing history, previous purchases, and similar behavior of other visitors to the site.
- Predictive analytics : anticipates user needs and preferences based on machine learning models to proactively suggest products.
- Chatbots and intelligent virtual assistants : These use NLP and reinforcement learning to interact with users and guide them through product discovery.
- Real-time personalization : which uses the power of AI to dynamically adapt content and recommendations based on the user's real-time behavior on the site.
Amazon's Rufus assistant Amazon recently tested a new shopping assistant based on generative AI built directly into product data sheets. This model, trained on Amazon's catalog and web data, suggests questions customers should ask to further explore the catalog and pushes products that illustrate the answers. Thanks to Generative AI, the assistant can answer complex questions, such as purchase criteria to consider or comparisons between different items. In this way, the tool facilitates product discovery and decision-making directly on the Amazon platform, with a conversational experience.
Machine learning applied to search engine
To go beyond the above understanding of the query, it is also possible to use machine learning algorithms with the e-commerce site's search engine to suggest products in the case of an unknown expression. Let's take a concrete example: if the site visitor types "chimichurri", the engine will be able to semantically understand that this is a sauce and will suggest others if it is not in the catalog. This reopening algorithm is an essential feature to avoid the "0 result" that drives visitors to another site.
These algorithms also open up new perspectives for personalization, which can be achieved at the product ranking level by taking into account :
- Seasonality: the engine suggests items according to the season. If a visitor enters "sweater" in summer, it will suggest light sweaters
- User preference: if the visitor searches for men's products, then shoes, the engine will suggest men's shoes.
- User profile : algorithms classify customers into different types of profiles, enabling personalized search results and recommendations, with customized listings right from the home page.
- Keyword Learning : the system learns to better understand the user's intent behind keywords. For example, for "jeans," it will prioritize pants over jackets based on best-sellers.
Another great advantage of AI in the front-end is its ability to constantly learn and adapt. By analyzing user behavior data and purchasing trends, AI can adjust e-merchandising strategies, refine product discovery algorithms, and improve chatbot responses. This continuous improvement process ensures constant optimization of the user experience and sales performance.
From search engine to answer engine thanks to AI: the example of Google
Search engines have evolved from a simple list of links to widget-based response elements. More recently, some Google searches have returned optimized snippets (called Featured Snippet) containing a concise answer generated by algorithms that analyze web page content to provide a precise answer to the query.
This transformation from search engine to answer engine reached a new stage with AI chatbots: by integrating Generative AI, Google announced a Search Generative Experience, which it later renamed AI Overviews at its annual conference. Gradually integrated into the SERPs, these AI Overviews offer, in addition to an answer generated from the sources found, the possibility of refining the question by switching to conversational mode with the user. By using AI, the engine is able to quickly understand the user's intent, even for complex and ambiguous queries.
For product searches, it also integrates customer reviews and nearby store locations for an even more personalized experience.
This new search experience (not yet announced in France) combines RAG technologies to retrieve current and relevant information and LLM to generate answers.
This evolution of the Google search engine brings new challenges for SEO. It will also mean that web users will become accustomed to obtaining information that is even more relevant and personalized to their queries, an advantage they will expect to find on the e-commerce sites to which the engine leads them.