Rakuten Super SALE Adds AI Concierge for Product Search

🔑 Enhanced Key Takeaways

•The AI concierge is a core component of Rakuten's broader 'Agentic Ecosystem' vision, aiming to provide hyper-personalized and interconnected experiences across its more than 70 diverse services.
•It leverages Rakuten's proprietary large language models (LLMs), including Rakuten AI 3.0, which is specifically optimized for Japanese language performance and boasts a claimed 90% lower inference cost compared to external frontier models like GPT-4o.
•Beyond internal product data, the AI concierge integrates general web search results to offer recommendations that consider real-world factors such as climate, popular trends, and the social landscape.
•The AI aims to mitigate 'choice paralysis' for users navigating Rakuten Ichiba's extensive catalog of approximately 500 million items by understanding latent user needs and lifestyle preferences beyond simple keyword matching.
•Rakuten's AI integration has demonstrated significant commercial benefits, contributing ¥25.5 billion in profit across its ecosystem in 2025 and leading to an approximate 26% increase in average order value for users engaging with the Rakuten Ichiba AI agent.

📊 Competitor Analysis▸ Show

Feature / Platform	Rakuten AI (Rakuten Ichiba)	Amazon Rufus	LINE and Yahoo! JAPAN's Agent i	General AI Shopping Assistants (e.g., Shopmate, Verloop, Tidio)
Core Functionality	Conversational product search, shopping strategy advice, personalized recommendations, intent understanding	Conversational product search, answers product questions, recommendations	AI agent for guiding shopping experiences	Personalized recommendations, customer support (FAQs, order tracking), generative AI conversations
Input Methods	Text, voice, image input	Conversational (text)	Not specified, likely text/voice	Text, sometimes voice
Data Integration	Leverages Rakuten's extensive ecosystem data (e-commerce, fintech, travel), integrates general web search results	Amazon's product catalog and customer data	LINE/Yahoo! JAPAN ecosystem data	Store's product catalog, FAQs, documentation, customer behavior data
Ecosystem Scope	Deeply integrated across Rakuten's 70+ services (e.g., Ichiba, Fashion, Travel, FinTech)	Primarily focused on Amazon's retail platform	Integrated within LINE and Yahoo! JAPAN's services	Often third-party solutions, integrate with platforms like Shopify, WordPress
Unique Selling Points	Addresses 'choice paralysis' with 500M items, Japanese-optimized LLM, agentic AI for proactive actions	Enhances traditional search and recommendation	Aims to assist with entire purchasing process	Automates support, increases agent productivity, customizable chat flows
Monetization/Impact	Contributed ¥25.5bn profit in 2025, 26% AOV increase for AI agent users	Enhances purchasing decisions, improves customer experience	Transforms shopping experience	Increases conversion rates, reduces support workload, boosts sales

🛠️ Technical Deep Dive

Rakuten AI 3.0 is a large language model utilizing a Mixture of Experts (MoE) architecture.
The model has approximately 700 billion parameters, with only about 40 billion parameters activating per token during inference.
It was trained on an in-house multi-node GPU cluster in a secure environment, ensuring all data remained internal.
Rakuten has developed custom language modeling work, including a specialized tokenizer for the Japanese language and culture.
The AI concierge employs semantic search technology to understand the meaning of user queries, significantly reducing instances of no-match searches.
Rakuten operates a centralized AI stack that encompasses data management, model development, APIs, and user experience integration, enabling efficient deployment across its diverse businesses.
The company develops both powerful foundational models for broad applications and cost-effective, specialized AI models tailored for specific purposes.
Rakuten's AI team comprises 1,000 personnel and utilizes thousands of Nvidia chips for its projects.

🔮 Future ImplicationsAI analysis grounded in cited sources

Rakuten's AI concierge will significantly increase customer engagement and loyalty across its ecosystem.

By providing hyper-personalized recommendations and seamless access to diverse services, the AI aims to anticipate user needs and create a more interconnected experience, fostering deeper engagement.

Rakuten will likely license its specialized AI capabilities, particularly its Japanese-optimized LLMs, to external businesses.

Rakuten's centralized AI stack and development of cost-efficient, domain-specific models, along with its 'Rakuten AI for Business' platform, position it to offer AI-as-a-Service, similar to hyperscalers.

The success of Rakuten's agentic AI will drive further adoption of conversational and image-based shopping interfaces across the Japanese e-commerce market.

As a major player, Rakuten's successful implementation of an AI concierge that understands intent and takes action will set a new standard for user experience, pushing competitors to follow suit.

⏳ Timeline

2022-09

Rakuten Ichiba celebrates its 25th anniversary, and Rakuten Super Sale its 10th year.

2024-03

Rakuten implements AI-powered semantic search on Rakuten Fashion and Rakuten Ichiba, reducing no-match searches.

2024-03

Rakuten releases its first high-performance Japanese large language models (LLMs), including Rakuten AI 7B.

2025-03

Rakuten's CEO highlights AI's contribution to operating income and the use of RMS AI Assistant (Beta) by merchants.

2025-07

Rakuten Group announces the full-scale launch of 'Rakuten AI,' unveiling its vision for an Agentic Ecosystem, with initial rollout in Rakuten Link and a web app (beta).

2025-12

Rakuten Ichiba launches the new agentic AI concierge in its mobile app.