The Tech ArchiveThe Tech ArchiveThe Tech Archive
Small BusinessMarketingDevelopers
ArticlesTopicsSeriesAbout

Get the practical AI brief

Verified, no-hype AI tips you can actually use - in your inbox. Free.

No spam. We verify what we send. Unsubscribe anytime.

The Tech ArchiveThe Tech Archive

The Tech Archive

AI news, analysis & explainers

AboutSmall BusinessMarketingDevelopersArticlesTopicsSeriesMethodologyAI DisclosureCorrections

© 2026 All rights reserved.

Back to home
0 readers reading
  1. Home
  2. Articles
  3. Artificial Intelligence
  4. Flipkart’s Agentic Shift: Why the E-Commerce Giant is Building Its Own LLMs

Contents

Flipkart’s Agentic Shift: Why the E-Commerce Giant is Building Its Own LLMs
Artificial Intelligence

Flipkart’s Agentic Shift: Why the E-Commerce Giant is Building Its Own LLMs

Flipkart is now generating 40% of its code with AI and building specialized e-commerce LLMs. Discover why general-purpose AI isn't enough for the scale of 2026.

Sham

Sham

AI Engineer & Founder, The Tech Archive

5 min read
1 views
July 3, 2026

Verdict: Flipkart’s transition to a "mixture of experts" model—combining frontier LLMs with proprietary, domain-specific models—is the new blueprint for enterprise AI. By automating 40% of its code generation and deploying 250+ specialized models, Flipkart is proving that the future of competitive advantage lies in owning the "secret sauce" of vertical AI rather than relying solely on general-purpose systems.

Last verified: July 3, 2026 · Core metric: 35-40% AI code generation · Model count: 250+ active deployments · Key player: Balaji Thiagarajan (CPTO, Flipkart)

Why general-purpose AI isn't enough for Flipkart

In a recent reveal by Chief Product and Technology Officer Balaji Thiagarajan, Flipkart confirmed it is moving away from a total reliance on general-purpose models like GPT-5 or Claude. While these "frontier" models are powerful, Thiagarajan noted that "general-purpose models won't be enough" to handle the nuanced, data-heavy requirements of a platform serving hundreds of millions of users.

Flipkart’s solution is a Mixture of Experts (MoE) architecture. This approach uses specialized models for distinct tasks—such as catalogue enrichment, product discovery, and seller support—while leveraging proprietary data that competitors cannot access. This strategy mirrors the broader shift we’ve seen in the rise of the loop engineer, where the focus shifts from prompting to designing autonomous systems.

The 40% code milestone: AI as a primary developer

Perhaps the most striking metric is that approximately 35-40% of Flipkart’s software code is now generated by AI tools. This isn't just about simple scripts; AI is now a core part of the engineering workflow.

This high level of automation allows Flipkart to scale its "agentic e-commerce platform" faster than traditional development cycles would permit. However, the company maintains that this is not a replacement for human judgment. Much like the 2026 reality for developers, Flipkart’s engineers are shifting toward roles of governance, content moderation, and "response fidelity" rather than manual syntax writing.

Key AI initiatives at Flipkart (2026)

Flipkart has already integrated GenAI across its entire ecosystem. Notable examples include:

Tool/Initiative Function Impact
Seller Lens Merchant Insight Platform Provides AI-driven growth recommendations to sellers.
Voice AI Agents Automated Seller Support Currently makes 90,000+ personalized calls monthly.
Conversational Shopping Customer Experience Uses proprietary LLMs for natural language product discovery.
Agentic Platform Infrastructure A framework for autonomous AI teams to handle supply chain and ops.

These initiatives are part of a broader trend where companies are building Agent Operating Systems to manage complex, multi-step business processes autonomously.

The strategy: Governance over immediate ROI

Despite the heavy investment required for custom LLM development, Flipkart is prioritizing governance and fidelity over immediate financial returns. The company measures success through business-critical metrics such as:

  • Conversion rates and customer engagement.
  • Basket sizes and demand forecasting accuracy.
  • Inventory performance and supply chain efficiency.

By focusing on the "fidelity" of AI responses and keeping humans in the loop for high-stakes decisions, Flipkart is insulating itself against the risks of labor arbitrage collapse currently affecting the wider Indian IT sector.

What this means for you

For business owners and developers, Flipkart’s "mixture of experts" strategy offers three clear takeaways:

  1. Vertical beats horizontal: Don't just use AI; build systems that use your unique business data as the training signal.
  2. Focus on the loop: Automation is most valuable when it handles the 40% "grind" of coding or operations, freeing your team for higher-order strategy.
  3. Governance is the new moat: As AI becomes a commodity, the trust you build through verified, accurate, and moderated AI outputs will be your primary differentiator.

FAQ

Q: Is Flipkart replacing its developers with AI?
A: No. While AI writes 40% of the code, Flipkart has actually been hiring senior leadership from companies like Amazon and Coupang to scale its AI governance and engineering capabilities. The role of the developer is evolving into one of system architecture and oversight.

Q: Why build proprietary LLMs instead of using OpenAI?
A: General-purpose models lack the specific context of e-commerce at Flipkart's scale. Proprietary models allow for better "response fidelity," lower latency for specific tasks, and the ability to leverage internal datasets securely.

Q: What is an "agentic e-commerce platform"?
A: It is a platform where AI agents don't just answer questions but take actions—like calling sellers for payment reminders or optimizing inventory levels—autonomously within defined guardrails.

Q: How does Flipkart measure AI success?
A: Instead of pure cost-savings (ROI), they look at conversion rates, average basket sizes, and customer engagement metrics.

Sources
  • Moneycontrol Interview with Balaji Thiagarajan (July 2026)
  • India Today Tech Analysis (July 3, 2026)
  • NewsBytes Science & Tech Report (July 2, 2026)
  • Flipkart Corporate Hiring & Strategy Updates (2026)
Updates & Corrections
  • 2026-07-03: Article published based on CPTO Balaji Thiagarajan's latest strategy reveals. Verified 35-40% code generation metric across three primary news sources.

Get the practical AI brief

Verified, no-hype AI tips you can actually use - in your inbox. Free.

No spam. We verify what we send. Unsubscribe anytime.

Discussion

0 comments
Sham

Sham

AI Engineer & Founder, The Tech Archive

AI engineer (Azure AI-102/AI-900). Writes practical, tested, hype-free guides on using AI for real work and small business at The Tech Archive.

Related Articles

View all
The $2.65 Billion Token Bill: Why Enterprise AI Agents Are Stalling at the Finish Line
Artificial Intelligence

The $2.65 Billion Token Bill: Why Enterprise AI Agents Are Stalling at the Finish Line

5 min
The End of the One-Time Chip Sale: How NVIDIA Now Earns from Every AI Token
Artificial Intelligence

The End of the One-Time Chip Sale: How NVIDIA Now Earns from Every AI Token

5 min
Hermes Agent v0.18 Review: The 'Judgement Release' Ends the AI Vibe Check
Artificial Intelligence

Hermes Agent v0.18 Review: The 'Judgement Release' Ends the AI Vibe Check

6 min
The 5-Layer Agentic Stack: How to Build Your Own Agent Operating System (2026 Guide)
Artificial Intelligence

The 5-Layer Agentic Stack: How to Build Your Own Agent Operating System (2026 Guide)

6 min
The New Meritocracy: Why Geography No Longer Limits Global Leadership in 2026
Artificial Intelligence

The New Meritocracy: Why Geography No Longer Limits Global Leadership in 2026

5 min
The 1% Rule: Why AI Fails at 99% of Trading Strategies (and How to Find the Winners)
Artificial Intelligence

The 1% Rule: Why AI Fails at 99% of Trading Strategies (and How to Find the Winners)

6 min