In search of a costume for her upcoming college reunion, Arti, a software analyst types a specific search term on her favourite e-commerce portal. The artificial intelligence embedded in the portal guides her by displaying different product listing pages (PLP) as per the search rankings. Further, on the product detail page, size and fit, similar products etc., are displayed. All these recommendations are personalised for Arti and take into consideration the intent of her shopping journey and other objectives like diversity, availability of products, sizes in inventory, etc. Arti finds the application effective as it helps her choose from the million styles on the portal and with a few filters is narrowed down to her personal requirements.
Broadly, the adoption of artificial intelligence for e-commerce revolves around harnessing a myriad combination of images and language models to present a mashup of choices for consumers. Different portals use different applications to achieve customer satisfaction. Myntra uses AI and generative AI-application, ChatGPT3.5 for query and intent understanding to show relevant and related products as collections. MyStylist is Myntra’s personal stylist who works on multiple machine language models developed in-house. MyFashionGPT caters to its return users and those buying from more than one category and Maya is Myntra’s own chatbot. “Our Homepage has personalised widget (tools used to drive functionality in e-commerce portals) ranking and within the widgets also personalised ad banner ranking. Our ranking models are highly advanced and use near real-time signals and explore-exploit algorithms,” says Raghu Krishnananda, CPTO, Myntra.
Flippi is Flipkart’s ChatGPT-powered shopping assistant. In addition to fine-tuned algorithms, says Mayur Datar, Chief Data Scientist, Flipkart, “We use regional language interfaces and voice capabilities to ensure seamlessness and user-friendliness in the shopping experience for customers from different geographies, and socio-economic segments.”
Amazon uses AI-ML-powered translation tools to break down language barriers and helps diverse consumers effortlessly navigate the platform and enjoy the experience. “Analysis of customer feedback and purchase patterns help us to refine our offerings and enhance the experience,” said Rajeev Rastogi, vice President, Machine Learning, Amazon. Among other applications, सह-AIis Amazon’s new digital shopping assistant that provides highly customised support to sellers and reduces their workload by simplifying time-consuming and complex steps such as registration, listing, and advertising support amongst others.
User engagement and conversion are two solid outcomes e-commerce portals aim to achieve through these applications. These, backed by search analytics and consumer behaviour analysis have certainly lived up to cater to the surge of pandemic-driven and post-pandemic online shoppers.
To satisfy this growing cohort, says Datar, “Personalisation becomes imperative. And personalisation is necessary on both the demand and supply side for Flipkart.” The portal’s brand mall mode is tailored for high-affluence users; Flipkart Samarth is for artisanal artefact lovers.
“At Amazon, personalisation serves as a vehicle to understand customer proficiency and offer them an adaptive experience; for low proficiency customers, onboarding tutorials and language options are presented for easy navigation. For high-proficiency customers, ads are upranked along with subscribe-and-save offers, and other sign-up widgets.”
Conversational chatbots help to drive personalisation closer to home. “When I am caught up in the maze of millions of styles and models, then I turn to the chatbot to narrow down my choices by making use of the context of my purchase,” says Arti. It is not rocket science but tech-savvy like her definitely enjoy an edge over other consumers, she adds.
Semantic searches, on the other hand, are driven by natural language processing algorithms (NLP) and used to figure out user intent support both textual searches and image searches such as allowing the users to take a photo and seek similar items.
“Leveraging state-of-the-art technology and large language models, Flipkart is bridging the syntactic gap in the way the users express their requirements and the way sellers describe their products. This is particularly crucial in a diverse country like India, where customers express their needs differently,” says Datar.
According to Statista, from 2022 to 2028, the Indian AI market is anticipated to expand at a CAGR of 30%. The metaverse will be where brands will provide a fantastic experiential environment for interaction and smooth commerce for consumers. Bespoke use cases will fuel breakthrough innovations. A new shelf-monitoring solution, a machine learning Wi-Fi and IoT camera-powered farm-to-fridge quality assurance system for fresh produce has set the pace for Amazon. It can detect the count of visible items of produce, and identify specific visual defects such as cuts, cracks, and pressure damage. This data is then used to alert store operators when items need to be removed from the shelf thus reducing food waste and also elevating customer shopping experience. “Providing consistently relevant and engaging experience is Amazon’s goal,” says Rastogi.
Soon, virtual shopping will become a livable experience for consumers. Krishnananda lists the many possibilities for GenAI in the fashion industry. “Helping customer care agents with faster issue resolution, automation of creatives for merchandising and notifications and enhanced virtual try-on experiences using AR/VR technology to enable users to virtually try on apparel and beauty products and accessories using live images.”
However, in creating a sustainable model, enterprises are not leaving out the concern for the privacy and security of customer data. Myntra processes all data in a PCI DSS-compliant environment. All customer credit and debit cards are tokenised and stored. Tokens are used for transaction processing. No card information is stored in the Myntra infrastructure. In fact, “AI-ML powered demand forecasting helps us prevent overstocking and unnecessary waste, promoting resource conservation,” says Rastogi.
Finally, Krishnananda, addresses the key question regarding AI robbing jobs. “The market will re-tool and re-train to accommodate. New roles such as Prompt Engineer could also emerge. In the end, it will be similar to the buggy-to-wagons-to-cars transformation-innovation,” he asserts.
