More and more companies are sharing some impressive numbers around how much of their coding they’ve managed to automate.
The latest to join that conversation is Meesho’s CEO Vidit Aatrey, who revealed that over 70% of the code at the Indian e-commerce platform is now AI-generated — a milestone that signals just how deeply artificial intelligence has been woven into the company’s operations.
Aatrey made the disclosure in an X post reflecting on Meesho’s journey and its AI-native direction. The post framed the company not as a retailer or a logistics player, but as a technology company — one that now ships code, products, and experiences at a pace that would be impossible without AI.
AI Is Reshaping How Meesho Builds
The 70% AI-generated code figure is striking on its own. But Aatrey paired it with another number that arguably matters more: platform experiments are up 2x year-on-year. That’s the downstream effect of AI-assisted development — the ability to move faster, test more hypotheses, and ship with greater confidence.
The most consequential AI deployment, however, isn’t in the codebase — it’s in how customers discover products. Over 75% of orders on Meesho now come not from search, but from personalised feeds powered by PRISM (Personalised Ranking and Intent Signal Module), an AI recommendation engine that infers consumer intent before the user has typed a single word. That’s a fundamentally different model from how most e-commerce platforms operate — and it’s working.
Meesho’s 70% figure puts it in notable company. Google’s Sundar Pichai had said in April 2025 that “well over 30%” of the company’s code was AI-generated; by October 2025, Google’s CFO had revised that figure to nearly half, just six months later. The number has since climbed to 75%, making Google one of the most cited benchmarks in this conversation. Microsoft’s Satya Nadella put his company at 20–30% around the same time. Closer to home, Inmobi CEO Naveen Tewari said in March 2025 that 80% of coding at the Indian ad-tech firm would be automated by end of 2025, and warned it would cost engineers their jobs. At the frontier labs, the numbers are even starker: Anthropic’s Dario Amodei estimated in September 2025 that 70–90% of code at Anthropic was being written by Claude, and by February 2026, Anthropic’s CPO said it had effectively reached 100%. What makes Meesho’s claim stand out isn’t just the number — it’s that a consumer e-commerce company focused on Tier 2 and Tier 3 India is sitting at 70%, alongside the world’s largest AI labs and tech giants.
Purpose-Built AI For Indian Conditions
Meesho’s AI push is distinctive because the company has built tools calibrated specifically for the Indian market. Vaani, its voice-based interface, brought in first-time users who had never completed an online purchase. This builds on Meesho’s earlier work replacing human support agents with AI bots, which cut support costs by 75% while simultaneously improving customer satisfaction scores. GeoIndia LLM was built to handle the messiness of Indian addresses — and it outperforms commercial geocoding systems. TrustMesh, the company’s fraud-detection layer, blocked 9 million high-risk transactions in FY26 alone.
Each of these is an unglamorous, infrastructure-level problem. Solving them at scale is what separates ambition from execution. Meesho already handles more daily orders than Amazon and Flipkart in India, making the efficiency gains from AI particularly meaningful at that scale. The narrowing losses are especially significant — this is a company that has historically prioritised growth over margins. AI is apparently helping it do both.