Over the last few months, Meesho joined the list of Indian e-commerce giants who have been progressively using AI to enhance user experience.
In an interview with AIM, Debdoot Mukherjee, chief data scientist and head of AI at Meesho, said, “There are many teams within Meesho who are successfully deploying LLMs in their applications.”
Meesho, after evaluating different generative AI solutions, stated that it chose Azure OpenAI to achieve sustainable cost optimisation of call centres.
At the same time, the company also adopted GitHub Copilot for its internal ecosystem of over 200 developers.
“Our partnership with Microsoft empowers us to deploy optimal generative AI solutions. This strategic move not only ensures seamless scalability during heightened demand but also enables sustainable cost optimisation, marking a significant step forward in enhancing customer support efficiency at Meesho,” Meesho CEO and co-founder Vidit Aatrey stated in a Microsoft blog.
The Bengaluru-based startup revealed in its FY 2023-24 earning calls that its revenue from operations grew by 33% to ₹7,615 crores.
“This success stems from driving efficiencies across multiple areas like logistics, as well as leveraging generative AI and machine learning for better discovery, improved in-app experience and round-the-clock customer support,” Meesho’s team stated.
Elaborating on the same, Mukherjee said, “There are at least a few top features or application areas where we are leveraging LLMs. One, of course, is customer support.”
He goes on to add that Meesho uses LLMs to scale human-intensive workflows, which become difficult to scale after a certain point.
“For instance, in customer support, when dealing with large volumes of inquiries, it becomes challenging to hire enough staff to manage and scale that workload effectively,” Mukherjee added.
Mukherjee also pointed out that maintaining relationships with sellers presents another opportunity for using LLMs. According to him, one of the essential steps in machine learning is creating annotated ground truth data which traditionally requires manual labeling by human annotators. This become a costly and time-consuming process.
“That, to a large extent, in many use cases, are getting automatically created and annotated by LLMs,” Mukherjee said.
Mukherjee mentioned that the second category of products or features that have greatly benefited from LLMs is a nuanced understanding of language, text, or images.
He affirmed that Meesho, as a company, has a variety of touch points and deals with a lot of Indic language text, which users type in vernacular words using the English keyboard pattern.
“So there’s a lot of transliterated text which we’re dealing with, and that kind of text is inherently very ambiguous,” Mukherjee said. “So, LLMs are helping in a very big way in understanding these queries and in validating the results that we deliver for them.”
At the NVIDIA AI Summit 2024, Mukherjee said that AI models are significantly powering the entire shopping experience at Meeshoo, which involves set-up journeys cataloguing, logistics, fraud detection and scaling fulfilment operations.
“We also use AI to recommend the right price points to be able to help our sellers get sales on our particular products. And we achieve this by mapping the products to similar products on the platform,” Mukherjee further said. “It’s fair to say that AI is riding the flight of the company in a very effective manner.”
Meesho is Not Alone
E-commerce giant Flipkart has also introduced its AI-powered search feature, Immerse, which offers a fusion of text and image search capabilities.
“[To] put it simply, imagine how an Indian shopper shops at stores. They start with something they like, then keep refining it. If they see a red dress but want it with long sleeves or in polka dots, they tell the shopkeeper, who helps them find what they want. This is precisely what Immerse aims to replicate in the realm of online shopping,” Flipkart’s senior product manager Aradhya Saxena said in a blog post.
Customers, however, don’t seem too satisfied. In a post on LinkedIn, one of the users discussed Flipkart’s AI-powered chatbot, Flippi, and shared that he will not use it as “there is no novelty attached to it.”
Another Reddit user expressed a similar concern, stating that when he randomly started asking questions and used some prompts, the chatbot swayed away from the main task and responded as ‘vanilla ChatGPT’.
“Our virtual assistant is one of the most significant examples of generative AI applications at Flipkart. Designed to mimic the behaviour of an in-store sales assistant, the virtual assistant provides real-time support to our customers by answering queries, offering product recommendations, and resolving issues during the shopping journey,” Mayur Datar, Flipkart’s chief data scientist, mentioned in a blog post.
He further explained how Flipkart is using generative AI in the multimodal search feature. “This tool allows our customers to search for products using a combination of text and images, improving the way users express their intent,” Datar added.
Moreover, Datar revealed that Flipkart is also using AI-generated images for merchandising, reducing dependence on traditional photo shoots and allowing sellers to showcase their products with high-quality visuals without needing professional resources.
Adding to the race, Amazon has also introduced Rufus, a shopping assistant powered by generative AI. With its initial rollout in the US earlier this year, Rufus is open to only select customers for now. “Customers in the US have already asked Rufus tens of millions of questions, and we’re excited to introduce it in India,” Amazon stated in a blog post.
Rufus is not new to this market. In 2023, Myntra collaborated with Microsoft to launch its AI-based personal style assistant, My Stylist. The feature recommends complete, shoppable looks tailored for fashion-forward customers. In the same year, they launched MyFashionGPT, a tool that enables users to search using natural language to give relevant suggestions.
Notably, the list of e-commerce businesses adopting generative AI continues to grow, with companies like Tata CliQ, BigBasket (now part of Tata Group), Nykaa and Lenskart embracing this approach.
The post Meesho Uses AI to Make Shopping Smarter appeared first on Analytics India Magazine.
