In this exclusive conversation with CIO&Leader, Anirban Nandi, VP and Head of AI Products & Analytics at Rakuten India, explains how the company’s GenAI strategy goes beyond pilots to create production-ready, customer-impacting features.
VP & Head of AI Products Anirban Nandi shares how GenAI is being embedded across Rakuten’s core ecosystem—moving from hype to hands-on value with governance, scale, and customer trust at its core.
CIO&Leader: How are you embedding GenAI into Rakuten India’s core product offerings today?
Anirban Nandi: GenAI is being integrated across multiple layers of our product stack, spanning both customer-facing solutions and internal operations. A great example is Rakuten SixthSense, our observability platform developed in India, which now delivers AI-driven insights across applications, APIs, and data pipelines. GenAI empowers teams to move swiftly from detection to resolution by offering contextual explanations and remediation suggestions. Essentially, observability enhanced with intelligence.
Across the Rakuten Ecosystem, we’re embedding AI in key use cases such as semantic search on the online marketplace Rakuten Ichiba to improve product discoverability, automating customer query responses in Rakuten Mobile, and enhancing campaign optimization in Rakuten Advertising through AI-driven insights. Internally, GenAI supports capabilities like auto-documentation, code generation, knowledge management, and sales enablement.
What makes this meaningful is not just the technology itself, but the scale of its deployment. These are not isolated experiments; they represent AI capabilities deeply integrated into production systems that enhance products serving millions of users.
CIO&Leader: What success metrics do you use to move AI initiatives from pilot to full-scale production?
Anirban Nandi: Our framework for advancing AI projects to production is grounded in three pillars:
Efficiency & Business Impact: We set quantifiable targets tied to specific outcomes. For example, we’ve reduced document creation time by up to 50%, accelerating sales email generation by up to 54%, and enabled faster responses to customer queries, reducing turnaround from hours to minutes.
User Adoption & Productivity: Early-stage pilots require consistent, daily usage by both technical and business teams. We track DAU/MAU ratios, task completion rates, and time saved, and use activity dashboards and template usage logs to measure both depth and breadth of engagement.
Systems Readiness & Governance: Scaling beyond pilots requires secure, enterprise-grade infrastructure. We evaluate integration into core platforms (e.g. embedding AI Analyst, Agent, Librarian across workflows), adherence to security and compliance standards, and governance around prompt management, version control, and observability. These are explicitly prioritized in Rakuten’s AI roadmap as foundational to enterprise readiness.
CIO&Leader: How do you balance personalization with data privacy and compliance in your AI systems?
Anirban Nandi: At Rakuten, we design for privacy from the onset. It’s not something we add later or tradeoff for performance. We train our models on aggregated behavioral signals and apply safeguards like differential privacy, strict data access controls, and internal red teaming exercises. This ensures we deliver relevant, high-quality experiences without compromising customer trust. We don’t view privacy as a barrier; it’s fundamental to how we build responsible AI systems.
CIO&Leader: Which AI-powered features have most improved customer engagement or conversion?
Anirban Nandi: Several features stand out:
- AI-powered recommendations have significantly increased basket size and repeat visits on our e-commerce flows.
- Semantic search has improved product discovery, especially for long-tail or ambiguous queries.
- Conversational support agents have reduced resolution times and improved satisfaction scores, while freeing up human teams for more detailed scenarios.
Together, these features deepen engagement by making customer experiences more intuitive and responsive.
CIO&Leader: How do you ensure cross-functional teams—product, analytics, engineering—work in sync on AI roadmaps?
Anirban Nandi: It starts with shared ownership. Every AI initiative at Rakuten India follows a triad alignment model where product, engineering, and data science co-own the goal from ideation to delivery.
We run joint planning cycles, maintain shared evaluation metrics that include both business KPIs and model performance, and foster a culture where hypotheses are tested collaboratively rather than simply handed off.
AI can’t live in silos, and we’ve institutionalised this spirit of collaboration as a core part of how we build.
CIO&Leader: What infrastructure investments (cloud, edge, GPU clusters) have been crucial to scaling AI at Rakuten India?
Anirban Nandi: Three areas have been critical:
- Cloud-native MLOps: Our pipelines leverage scalable, containerized systems for efficient model training, tuning, and deployment.
- GPU clusters: Dedicated GPU infrastructure, both on-premise and via cloud partners, powers high-performance GenAI training and inference.
- Data infrastructure: Unified data lakes, real-time ingestion, and annotation platforms ensure our models are built on high-quality, relevant data.
Scalable infrastructure is the difference between a prototype and a platform.
CIO&Leader: How do you cultivate an AI-first culture and build talent within your teams?
Anirban Nandi: Rakuten’s AI-nization initiative, the strategic and responsible infusion of AI across every layer of the organization, shapes how we build talent and embed AI into our day-to-day mindset, not just our technology.
At Rakuten India, cultivating an AI-first culture starts with access and ownership. We ensure every team, from product and engineering to operations and design, is equipped with the tools, training, and autonomy to explore AI in their workflows. Internal enablement programs, GenAI learning paths, and AI literacy workshops are available to all employees, regardless of role or function.
We also run cross-functional hackathons, and “build to learn” initiatives, where teams prototype real solutions to real business problems often leading to production-ready tools. Our engineers and product managers collaborate directly with Rakuten’s global AI teams, gaining exposure to advanced LLM architectures, prompt engineering, and multi-agent systems.
Importantly, we’ve integrated AI thinking into leadership development. Managers are trained to spot AI opportunities, evaluate ROI, and champion responsible deployment, ensuring cultural change is driven both bottom-up and top-down.
CIO&Leader: Where will Rakuten India’s AI strategy focus in the next 12–18 months—especially around GenAI or autonomous agents?
Anirban Nandi: Key focus areas for Rakuten India’s AI strategy includes:
- Agent orchestration: Building coordinated, task-oriented AI agents capable of handling multi-step workflows.
- RAG pipelines: Integrating retrieval-augmented generation with internal documentation and prompt libraries.
- Governance tooling & observability: Embedding usage dashboards and security/RAG controls to ensure trust and compliance.
- Localized LLMs: Leveraging models like Rakuten AI 2.0 for linguistic customization while maintaining efficiency.