E-commerce firms turn to AI to solve for Bharat during festive sales

Festive sales across e-commerce platforms are just around the corner, and artificial intelligence (AI) is being woven into customers’ shopping journeys.

From spotting trends before they become popular and predicting what consumers are likely to buy to personalising recommendations, changing advertising campaigns in real time, and helping deliver orders to the right address – AI is doing a bulk of the work behind the scenes.

Shoppers, too, are becoming more comfortable with AI being a part of their buying journey.


According to InMobi Advertising’s The Marketer’s Guide to India’s Festive Season 2026, 80% (four in five) of the shoppers want to use an AI shopping assistant this season and are interested in AI-powered features such as smart search suggestions, virtual product try-ons, complete-the-look recommendations and voice or chat assistants.

Pratik Kumar, Meesho’s head of engineering, told ET AI their AI-led TrendPulse system analyses around 400 million search queries and identifies about 65 emerging shopping trends daily.

“TrendPulse analyses 400 million searches daily to spot emerging shopping moments, often up to 60 days before they peak. With 97% of these moments having a regional or hyper-local dimension, it can identify signals from Hariyali Teej and Basant Panchami to fast-moving trends such as Kashmiri bangles,” he said.

During this year’s festive season, where 180-185 million Indians are expected to shop online as per market research firm Redseer, trend identification can become important, when demand changes quickly, and regional preferences vary.

“Everyone knows India buys phones and gold at Diwali. The real question is who starts buying what this week, and how much they’ll stretch,” Siddharth Kelkar, managing director, India/MENA and performance business at AnyMind Group, told ET AI.

AnyMind provides software and platforms for marketing, e-commerce, and digital transformation.

AI and the search-decide problem

Ecommerce companies are embedding AI on their platforms to better engage with customers during their shopping journey.

Walmart-owned Flipkart is using AI to combine search, conversations, voice, visual inputs, and content feeds to move beyond keyword-led product searches that can understand the intent of a customer, ask follow-up questions, and suggest recommendations.

Rufus, a shopping assistant for Flipkart’s rival Amazon, can answer questions about products, reviews and pricing; AI-generated review summaries help customers quickly understand positive and negative feedback.

Akshay Sahi, vice president, prime and customer fulfilment, Amazon India, told ET AI that the platform has introduced price history on product pages for shoppers to see how the price of a product has moved over time and judge the festive deals based on that.

”AI is embedded across the shopping experience. On product detail pages, customers can access review summaries that highlight positive and negative feedback, helping them make more informed decisions. At the back end, we use AI-powered systems to help curate the selection of products that are most relevant for Prime members,” he added.

Myntra’s chief product officer (CPO) Lakshminarayan Swaminathan said the company is using AI in a similar way across its fashion ecosystem. Its contextual styling tools suggest outfit pairings and its size-and-fit intelligence covers around 85% of its apparel portfolio.

“On the fulfillment side, automated capacity planning has reduced scenario planning from two days to about an hour, while real-time trend signals are helping us anticipate demand more dynamically,” he told ET AI.

What’s in it for brands?

Kelkar said brands are combining real-time intent, contextual behaviour, seasonal patterns, price and promotional signals to understand the buying preferences of the consumers. AI is making festive targeting, which has traditionally relied on broad categorisation, granular by helping brands identify behavioural intent and purchase propensity.

Kelkar added: “The other shift is that planning is becoming less static. AI agents can create an automated learning loop, where marketers can feed new signals for optimisation, instead of waiting for the next planning cycle to make a change.”

Rajiv Dingra, founder and CEO, agentic AI full-funnel agency ReBid, believes AI will move from recommending decisions to actually taking them.

Dingra told ET AI that agentic systems can potentially reduce the time for traditional campaign optimisation from hours to minutes. An AI agent could spot a fall in return on ad spend, figure out that a particular creative is suffering from fatigue, reduce its budget, move spending towards stronger audiences and then monitor whether the change worked.

“The shift is essentially from asking, ‘Who is my target audience?’ to ask, ‘Who is likely to buy this product in the next few days?,” Dingra said.

But brands are still wary of handing over complete control. Large budget changes, sensitive creative, pricing claims and brand-safety decisions are areas where human approval remains important.

“The model we believe will dominate is human-on-the-loop rather than human-out-of-the-loop,” Dingra said.

Solving for Bharat

The challenge for e-commerce companies in India during such sales is the diversity in the customer base. What sells in Mumbai may not do well in Lucknow or Guwahati. Problems arising from variations in languages used during search, including regional languages, or how people describe what they want need solutions to maintain customer stickiness.

Meesho claimed that about 97% of the shopping moments identified by its TrendPulse have a regional or hyperlocal dimension. Its PRISM, Meesho said, the system uses more than 100 AI ranking models to personalise product discovery, with more than 75% of orders coming from personalised feeds.

”Today, over 75 percent of the orders originate from AI-driven personalised feeds, connecting emerging demand with relevant products for individual shoppers. This intelligence extends across the commerce journey, helping Meesho respond to the complexity of festive demand at scale. PRISM-driven improvements have contributed to approximately 15% higher conversion and reduced the time for new listings to gain traction by about 30%,” Kumar added.

In addition, Meesho also claims that its GeoIndia LLM has improved geocoding accuracy by around 20 percentage points in shipping-heavy lanes and, together with its network intelligence system, reduced last-mile misroutes by more than 50%.

”Combined with Meesho’s Network Intelligence System, it helps reduce last-mile misroutes by more than 50%. Behind this scale, BharatMLStack enables Meesho to run AI workloads at 60–70% lower costs than equivalent cloud services, becoming particularly important as inference volumes can rise up to three times during peak sale days. Chorus, Meesho’s AI-powered customer support system, resolves 62% of customer queries, helping absorb the sharp increase in support needs during festive peaks,” Kumar added.

Flipkart, too, has a delivery intelligence system that uses AI-powered geocoding and last-mile orchestration to improve delivery speed and accuracy across its network.