Meesho bets on AI to get Bharat to shop more

As Meesho reported strong user growth with its annual transacting users increasing 33% year-on- year from 199 million in FY25 to 265 million in FY26, the company is betting on its tech stack investments to fuel growth through deeper penetration of smartphone users, especially in Bharat.   

This comes at a time when 70% of its code is AI-generated and its adoption across the Software Development Lifecycle in code and test case generation, code reviews, production monitoring and deployment fixes are on the up. “Platform experiments in Q4FY26 were up 2.0x versus the same quarter last year, and the compounding gains are showing up in platform reliability and quality,” CEO and founder Vidit Aatrey said in his letter to shareholders.  

Outlining some of its engineering initiatives, Meesho said that its suite of proprietary AI systems not just compounded efficiency across every layer of its operations, but were also making small-ticket e-commerce economically viable at a population scale.

For instance, the GeoIndia LLM, trained on millions of real delivery, traces across thousands of pin codes, converts vernacular and landmark-based addresses into precise coordinates, thus improving geocoding accuracy by 20 percentage points and cutting misroute-related costs by 5%. Meanwhile TrustMesh, its deep learning integrity model, blocked around 9 million high-risk transactions, restricted 2 million consumers and 62,000 sellers, and reduced return-to-origin rates by over 10% in FY26.  

Also Chorus, Meesho's voice and chat agentic platform, resolved 19 million consumer calls without human intervention in FY26, cut customer support costs by 23%, handling up to 300,000 seller calls per day—driving ad adoption and improving seller participation in promotional events.   

With Meesho Labs—the AI focussed research and innovation unit of the company—building BharatMLStack, Meesho's in-house ML platform, which the company says runs at 60–70% lower cost than equivalent cloud services, it has already substantial amounts in terms of investments on the tech side.  

According to Meesho's outlay plan for IPO proceeds, the company plans to spend ₹610 crore each in FY27 and FY28, and about ₹170 crore in the first quarter of FY29 for investment in cloud infrastructure. At the same time, the company’s subsidiary, Meesho Technologies Private Limited, plans to deploy ₹240 crore each in FY27 and 28 for AI talent hiring and development. 

Its latest bet on AI voice support for deeper customer engagement in its test sample has seen the company improve its conversion by 22% for those who adopted it. This bet gains significance given that though the company has seen its order frequency rise by 9% in FY26 compared to 7% last year, the average order value has seen a dip of 8% and 3% over the last two fiscals.

This March, in a conversation with Fortune India, co-founder and CTO Sanjeev Barnwal had said that it would take some time to fully roll out the AI voice support in other Indic languages; the feature is primarily available in Hindi. He did not rule out collaborating with other companies that are building LLMs in Indic languages.  

Elaborating on the how the company’s GenAI journey over the last two years has translated both in terms of efficiency and impacting the bottom line, Barnwal had said that while automating both voice- and chat-based customer support led to efficiency gains and reduced costs, there was also an increase in productivity within engineering teams. Now, the AI voice interactive shopping assistant would lilely aid the top line going forward. “We also use Generative AI today in our search, under the hood. So, when you search for something on Meesho, it's not very apparent but behind the scenes, we use Generative AI to understand that query... [it] is much better than traditional AI models helping in conversion,” he said, adding that is a differentiator compared to other platforms .  

Meesho’s shares rose by as much as 7.6% intraday on May 7, after the company reported that losses had narrowed losses in Q4FY26 to ₹166 crore against ₹1,391 crore a year earlier. Analysts at JM financials noted, “In FY26, Meesho reported negative FCF of ₹6.33 billion versus a positive FCF of ₹5.91 billion in FY25, due to the disruptions. We see NMV trends normalising to ~34% YoY in FY27 on account of a tough base (NMV was up ~39% in FY26) and expect the pace of sequential Adj. EBITDA margin expansion to slow down as a reasonable proportion of inefficiencies were likely addressed in 4QFY26 itself. Moreover, the stock currently trades at ~46x FY29E Adj EBITDA, one of the highest multiples in our Internet coverage, leaving little room for any misses.” 

On the Q4FY26 numbers,  analysts at Univest observed that though the improvement in contribution margin trajectory, combined with the sustained 33% growth in annual transacting users, builds a medium-term investment case at current levels, that, however, hinges on execution, whether advertising revenue meets expectations over the next two to three quarters.