Flipkart bets big on generative AI to improve customer, seller experience

Walmart-owned e-commerce major Flipkart is betting big on artificial intelligence (AI), especially generative AI applications, to improve the experience of its customer and seller ecosystems. The company is leveraging this technology across the board in all areas of the business to improve efficiency, an executive at the firm said.

“E-commerce is like having many sectors under a single roof…AI is applied in every aspect of e-commerce, from the supply chain, to preventing fraud, to advertising, to improving the search experience,” said Mayur Datar, chief data scientist, Flipkart.

Datar expects that, as the company builds scale, it will be able to improve on the returns on investment (ROI) it is witnessing from its AI bets.

“Today, it (technology) is expensive, the underlying models are costly. We use a lot of in-house models but these also include training and hosting which requires custom hardware like GPUs, which are expensive. However, costs are coming down and this will eventually be net return on investment (ROI) positive. Even today, in many cases, it is net ROI positive,” Datar said.

He added that Flipkart is leveraging generative AI tools to improve its customer and seller experiences.

On the customer side, Datar said that Flipkart is implementing generative AI use cases such as creating its own virtual assistant, dubbed Flippi. Although still in its early stages, the “offline style conversational shopping interface” provides customers with additional information about products, offers, return policies, etc.

To improve its search functions, the company is experimenting with applications such as multimodal search – which provides improved search through multimodal inputs like text, images, and speech. It is also working on improved intent understanding, which allows for better interpretation of search keywords. For instance, converting “bacho ke liye dress” to “dress for kids”.

Flipkart’s generative AI tools on the seller side, Datar said, include tools such as automated catalogue shoots – which enables sellers to convert images of products to what they would look like on a model using generative AI. The company is also using AI for stock image generation and extracting product attributes in an automated manner.

Datar, however, said that the application of AI is not just limited to the customer and seller ecosystem. The e-commerce major is also leveraging this technology in areas like supply chain optimisation to address challenges like network optimisation, inventory placement, and geocoding and address autosuggest.

The firm is also seeing a big opportunity in advertising. Datar claimed that, using AI, the company is estimating factors such as the likelihood of customers clicking on advertisements and the probability of those clicks ending up in purchases.

“We are doing this at an enormous scale. Millions of estimations are done during the day, in a fast manner. And the model keeps updating itself based on how the customers are reacting,” he said.

Datar, however, emphasised that the company is making sure to follow responsible implementation of AI in a manner that does not harm its customers.

First Published: Apr 25 2024 | 6:33 PM IST