40% of Code AI-Generated: Flipkart Surges Ahead by Building Specialized E-commerce LLMs in Digital Race

40% of Code AI-Generated: Flipkart Surges Ahead by Building Specialized E-commerce LLMs in Digital Race​

Flipkart is rapidly accelerating its commitment to artificial intelligence (AI), revealing that nearly 35 to 40 percent of its software code is now generated by AI tools. The e-commerce giant is not merely utilizing existing AI assistants but is actively building specialized, proprietary Large Language Models (LLMs) designed specifically for the complex needs of the e-commerce ecosystem.

Chief Product and Technology Officer Balaji Thiagarajan stated that Flipkart has deployed over 250 such models across its platform. This massive ramp-up aims to integrate AI deeply into customer experiences, seller services, and core engineering processes within the Walmart-owned company.

Building an Agentic E-commerce Platform​

Flipkart is moving towards what it terms an "agentic e-commerce platform." This platform relies on a strategic combination of state-of-the-art frontier AI models alongside the specialized LLMs that Flipkart itself has developed.

Thiagarajan emphasized that this proprietary development is key to their differentiation in the market. He noted, "The future of AI will be a mixture of experts... many of the other experts will be models we build ourselves." This bespoke approach, driven by Flipkart's unique data and engineering capabilities, ensures functional specialization for specific e-commerce tasks.

Scaling AI Across Operations Beyond Coding​

Flipkart’s use of generative AI extends far beyond mere coding efficiency. The company is embedding these advanced systems across crucial business functions, including product discovery, conversational shopping, seller tooling, and catalogue creation.

AI now powers personalized shopping feeds for customers and features in the Seller Lens platform. This dedicated tool provides merchants with an AI-powered avenue to manage and scale their businesses on Flipkart.

The deployment of AI voice agents is particularly significant in seller operations. These automated systems currently make approximately 90,000 personalized calls each month to sellers. These calls are used for providing payment reminders, sharing operational updates, and giving business recommendations. The company projects that these volumes will soon scale into the hundreds of thousands, eventually reaching millions.

Governance and Metrics Drive AI Investment​

While generative AI technologies remain resource-intensive, Thiagarajan revealed that Flipkart is intentionally prioritizing robust governance over immediate financial returns in its AI initiatives. They are currently operating in an "investment mode" regarding AI.

A major focus area is therefore AI governance. This includes establishing strong systems for content moderation, ensuring response fidelity, and implementing human-in-the-loop protocols alongside reinforcement learning processes.

Instead of solely tracking cost savings, Flipkart measures the business impact of its AI through metrics such as customer engagement, average basket size, conversions, demand forecasting accuracy, and inventory turnover rates. The implementation has already shown positive effects in improved product discovery and increased average basket values.

The Intensifying AI Race Among Giants​

Flipkart's aggressive push comes amid a rapidly escalating race across the industry to embed artificial intelligence into core business processes. This acceleration is mirrored by major competitors.

Amazon recently announced an $13 billion investment into India through 2030, targeting its cloud and AI infrastructure. This expansion brings Amazon’s total planned investment in the country to $48 billion as it scales its quick commerce and e-commerce operations.

Meesho has also significantly ramped up its AI integration, stating that over 70 percent of its code is now AI-generated. AI is becoming foundational to their software development, logistics, advertising, and customer support functions.

Thiagarajan concluded by underscoring the long-term value of specialized systems. He stated clearly that general-purpose models will ultimately be insufficient for complex business tasks. The "secret sauce," he asserted, is building custom models tailored to specific organizational requirements.