Meesho is going all in on artificial intelligence. It sits at the core of how the platform works, from shopping and deliveries to ads and fraud checks.
According to its latest IPO filing, AI and machine learning now power almost every interaction on Meesho. When users search for products, browse their feeds, place an order or contact support, there’s an algorithm working in the background. The same applies to sellers listing products, running ads or managing prices.
To drive this, Meesho has set up Meesho AI Labs, its in-house AI team. The focus here is on building models for Indian users, including small and large language models, agent-based AI for shopping and post-order journeys, stronger risk systems and more automated ad optimisation (see infographic). The company follows a simple approach: test new AI ideas, measure impact, and scale only what works.
.png)
source: Meesho.com updated Red Herring prospectus ; Infographic generated using my own prompt.
Behind the scenes, Meesho runs its own machine learning platform called BharatMLStack. It’s built to handle India’s scale while keeping costs low and response times fast. In FY25 alone, the platform processed close to 2 petabytes of data every day. These systems power real-time recommendations, pricing decisions, geo-location mapping and fraud detection.
One of the more practical uses of AI is in solving India’s address problem. Meesho has built a custom GeoIndia language model that turns messy, unstructured addresses into accurate map coordinates. This has helped improve last-mile delivery and cut fulfilment costs.
For shoppers, AI shows up as more personalised feeds and better search. Users can search using text, images or voice, even in local languages or with spelling mistakes. AI-powered chat and voice bots now handle a large share of customer queries. In FY25, they resolved about half of all support requests, and this crossed 60% in the June 2025 quarter, reducing wait times and support costs.
Sellers also benefit from AI tools. These help with onboarding, catalogue creation, pricing insights and targeted ads. Meesho says its AI-driven ad systems delivered an 8.6x return on ad spend in FY25, which jumped to 18.3x in the June quarter, as targeting became more precise.
Logistics is another area where AI plays a big role. Meesho’s logistics arm, Valmo, uses dynamic routing models instead of fixed shipping routes. These systems look at the network in real time, select the best delivery partners, predict disruptions and reroute shipments during capacity or weather issues. AI also helps read complex delivery addresses and improve delivery success ( see infographic below)
The Logistics System
Trust and safety remain critical. Under Project Vishwas, AI systems track account misuse, fake GPS signals, bot activity and transaction fraud. Project Suraksha uses computer vision and language models to spot counterfeit or infringing listings, backed by continuous monitoring of product quality through customer reviews.
Meesho is also using generative AI internally. Engineers use it to speed up coding and product releases, while marketing and product teams rely on AI to create images and videos for campaigns. The company is even testing whether some AI-powered support tools can be offered to external partners.
All of this runs on massive data. Meesho’s AI systems process over 4.3 billion data points every day. Even as transaction volumes grow, the company says its infrastructure costs are rising much more slowly, showing how AI is helping improve efficiency.
People remain a big part of the plan. As of June 2025, Meesho had 155 AI and ML specialists, and more than half its workforce is in tech roles. A portion of the IPO funds will go towards hiring and retaining AI talent and strengthening long-term AI capabilities.
In summary, it should be interesting how the company ramps up and integrates AI as it scales up. A lot rides on successful implementation and adoption!
No comments:
Post a Comment