Meesho on Thursday said its proprietary AI system PRISM — Personalised Ranking & Intent Signal Module — now powers more than 75 per cent of all orders placed on the platform, as the Bengaluru-based e-commerce company disclosed fresh scale metrics for the technology it has been building for India’s next wave of online shoppers.
The platform, which reported 264 million annual transacting users and 717 million placed orders in the fourth quarter of FY26, said PRISM processes close to 100 million inferences per second during peak sale events and runs over 6 trillion inferences daily. The system draws on a network of more than 100 AI ranking models trained on over 400 trillion input signals.
Unlike conventional search-bar driven e-commerce, PRISM is designed around browsing and discovery behaviour — what Meesho describes as the dominant shopping pattern among first-time and vernacular internet users. The system supports more than 10 languages including Hindi, Bengali, Tamil, Telugu, and Odia, and is underpinned by BharatMLStack, Meesho’s in-house machine learning infrastructure built to run high-throughput workloads at lower costs than standard cloud setups.
The company also highlighted Trendpulse, a large language model-powered engine within PRISM that tracks emerging demand patterns across regions and consumer clusters to surface relevant products ahead of formalised search intent.
Debdoot Mukherjee, Chief Data Scientist and Head of AI at Meesho, said the next hundred million Indians coming online would not search but discover, calling it the most significant shift in consumer internet India has seen.
Shares of Meesho Limited, listed on the NSE since December 2025 and part of the Nifty Smallcap 100 index, fell 5.01 per cent to ₹166.62 by mid-afternoon on June 4, dragging the stock to near the lower end of its ₹125.56–₹254.40 fifty-two-week range. Total market capitalisation stood at approximately ₹76,620 crore.