Rakuten Launches AI Assistant for Enterprise

Rakuten, Japan's largest e-commerce conglomerate, has launched an enterprise-grade AI assistant purpose-built for Japanese business workflows. The new product, dubbed Rakuten AI for Business, represents one of the most significant moves by an Asian tech giant to challenge Western enterprise AI platforms like Microsoft Copilot and Google Duet AI in the rapidly expanding Asia-Pacific market.

The assistant integrates natively with Japanese corporate communication standards, document approval chains, and compliance frameworks — areas where Western AI tools have historically struggled to gain traction. Rakuten estimates the platform could save enterprise clients an average of 12 hours per employee per month by automating routine administrative tasks unique to Japanese business culture.

Key Facts at a Glance

• Product name: Rakuten AI for Business, launching initially in Japan with plans for broader APAC rollout
• Target market: Mid-to-large enterprises with 500+ employees operating in Japan
• Pricing: Starts at approximately $25 per user per month, undercutting Microsoft Copilot's $30 tier
• Language model: Built on Rakuten's proprietary large language model, fine-tuned on Japanese business corpora
• Key differentiator: Native integration with Japanese workflow standards including ringi (consensus-based approval), nemawashi (pre-meeting coordination), and hanko (digital seal) systems
• Availability: Enterprise beta begins Q3 2025, with general availability expected by early 2026

Why Japanese Business Workflows Need Specialized AI

Japanese corporate culture operates on a set of deeply embedded administrative processes that have no direct equivalent in Western business environments. The ringi system, for example, requires documents to pass through multiple layers of approval before any decision is finalized — a process that can involve dozens of stakeholders and take weeks to complete.

Western enterprise AI tools like Microsoft Copilot and Salesforce Einstein treat approval workflows as linear processes. Rakuten AI for Business, by contrast, understands the circular, consensus-driven nature of Japanese decision-making and can draft, route, and track ringi documents automatically.

The platform also handles keigo — the complex system of Japanese honorific language that changes based on the relative status of sender and recipient. Getting honorifics wrong in a business email can damage relationships and derail deals. Rakuten claims its model achieves 97% accuracy in keigo usage, compared to roughly 78% for GPT-4o when handling similar Japanese business correspondence.

Technical Architecture Powering the Platform

Rakuten AI for Business runs on the company's proprietary Rakuten LLM, a family of large language models trained specifically on Japanese-language data. Unlike many Japanese AI startups that fine-tune Meta's Llama or other open-source Western models, Rakuten built its foundation model from scratch using a training corpus of over 1.2 trillion Japanese tokens.

The architecture includes several notable components:

• Workflow Engine: A rule-based orchestration layer that maps AI outputs to specific business process templates
• Context Memory Module: Maintains persistent memory of organizational hierarchies, project histories, and communication preferences across sessions
• Compliance Gateway: Automatically screens all AI-generated content against Japan's Act on Protection of Personal Information (APPI) and industry-specific regulations
• Multi-Modal Input: Supports text, voice, and scanned document inputs — critical for organizations still transitioning from paper-based processes
• API Integration Layer: Connects with popular Japanese enterprise tools including Cybozu kintone, Sansan, and Money Forward, as well as global platforms like Slack and SAP

The model itself reportedly contains 70 billion parameters and was trained using Rakuten's own GPU cluster, which the company expanded significantly in late 2024 with a reported $420 million infrastructure investment.

Rakuten Takes Aim at a $45 Billion Market

The Asia-Pacific enterprise AI market is projected to reach $45.1 billion by 2028, according to IDC, growing at a compound annual rate of 28.4%. Japan alone accounts for roughly $8.7 billion of that forecast, making it the second-largest national market in the region after China.

Despite this massive opportunity, adoption of enterprise AI tools in Japan has lagged behind the United States and Europe. A 2024 survey by Japan's Ministry of Economy, Trade and Industry found that only 23% of Japanese enterprises had deployed generative AI tools in production, compared to 48% in the U.S. and 39% in the EU.

The primary barrier is not technological reluctance but rather a poor fit between available tools and local work practices. Rakuten's CEO Hiroshi Mikitani has publicly argued that this gap represents a 'trillion-yen opportunity' for companies that can bridge Western AI capabilities with Japanese operational realities.

Rakuten is not alone in pursuing this thesis. NEC Corporation launched its own enterprise AI suite in early 2025, and Preferred Networks has been expanding its AI consulting services to large Japanese manufacturers. However, Rakuten's advantage lies in its existing ecosystem of over 1.6 million merchant partners and its deep integration with Japan's digital commerce infrastructure.

How It Compares to Western Enterprise AI Platforms

The most direct comparison for Rakuten AI for Business is Microsoft 365 Copilot, which dominates the global enterprise AI assistant market with an estimated 40% market share among Fortune 500 companies. However, Copilot's Japanese-language capabilities have drawn criticism from enterprise users in Japan.

Key differences include:

• Language precision: Rakuten's model is trained natively on Japanese, while Copilot relies on multilingual GPT-4o with Japanese as a secondary language
• Workflow integration: Copilot excels in Microsoft-centric environments but lacks connectors for Japan-specific enterprise tools
• Cultural context: Rakuten's assistant understands seasonal business customs (such as aisatsu greeting protocols and year-end oseibo gift coordination) that Western tools ignore entirely
• Data residency: All data processing occurs within Japan, meeting strict domestic data sovereignty requirements that some multinational AI providers cannot guarantee
• Pricing: At $25 per user per month, Rakuten undercuts Copilot by approximately 17%

That said, Microsoft retains advantages in global interoperability, English-language performance, and breadth of third-party integrations. For multinational companies operating across both Japanese and Western markets, a hybrid approach may prove most practical.

What This Means for Global Enterprise AI Strategy

Rakuten's launch signals a broader trend that enterprise AI leaders worldwide should watch closely: the localization imperative. As generative AI moves from experimental pilots to mission-critical deployments, organizations are discovering that one-size-fits-all AI assistants often fail to capture the nuances of local business practices.

This pattern is not unique to Japan. Similar dynamics are emerging in Germany, where AI tools must navigate complex works council regulations, and in the Middle East, where Arabic-language AI assistants need to handle right-to-left text and Islamic finance compliance simultaneously.

For Western AI companies like Microsoft, Google, and Salesforce, Rakuten's move underscores the risk of leaving localization gaps that regional competitors can exploit. The enterprise AI market may ultimately fragment along cultural and regulatory lines, with global platforms serving as infrastructure layers while local champions handle the 'last mile' of workflow integration.

Looking Ahead: Expansion Plans and Market Impact

Rakuten has outlined an aggressive expansion roadmap. Following the Japan launch, the company plans to adapt the platform for South Korea and Taiwan by mid-2026, followed by Southeast Asian markets including Singapore, Thailand, and Indonesia by 2027.

Each market adaptation will require significant model fine-tuning and workflow customization, but Rakuten believes its experience in Japan provides a replicable playbook. The company has also hinted at potential partnerships with local telecommunications providers to bundle the AI assistant with enterprise connectivity packages.

If Rakuten succeeds, it could establish a new category of culture-native enterprise AI — tools designed from the ground up for specific business cultures rather than translated from English-language originals. This would represent a meaningful challenge to the assumption that Silicon Valley's AI platforms will inevitably dominate global enterprise markets.

The stakes are high. Enterprise AI is quickly becoming the primary battleground for the next generation of business software, and Rakuten's bet on cultural specificity could prove to be either a masterstroke of market positioning or an expensive lesson in the limits of localization. Either way, it is a story that every enterprise technology leader should be following closely.