Financial institutions and AI share an uncomfortable relationship. Banks deal in data that is simultaneously among the most sensitive in existence — loan histories, creditworthiness assessments, transaction records, personal financial situations — and most valuable for AI training and analysis. The regulatory obligations that govern how that data can be stored, transmitted, and processed have made banks among the slowest large-enterprise categories to adopt AI productively, even as their operational workflows — credit assessment, document processing, compliance checking, customer inquiry handling — are among the most amenable to AI automation. Fujitsu’s answer to that tension is a platform designed specifically so that the AI and the data never have to meet outside the bank’s own walls.
On July 28, 2026, Fujitsu Limited announced it is initiating development of the Uvance for Finance AI Transformation Platform, a dedicated AI system for regional and international banks. Development formally begins August 1. The platform is built on the Fujitsu Kozuchi Enterprise AI Factory and integrates the Takane large language model — Fujitsu’s proprietary model co-developed with Cohere Inc., trained on financial sector legal frameworks, banking customs, and specialised terminology — alongside multi-AI agents for autonomous task execution and guardrail technology for security compliance.
Key Developments
- Fujitsu announced on July 28, 2026 the initiation of development of its Uvance for Finance AI Transformation Platform, designed specifically for regional and global banks, with development formally beginning August 1, 2026.
- The platform is built on the Fujitsu Kozuchi Enterprise AI Factory and integrates Takane — Fujitsu’s proprietary LLM co-developed with Cohere Inc. — which is trained on financial legal frameworks, banking customs, and sector-specific terminology.
- Multi-AI agents handle loan screening, document generation, and customer inquiries autonomously, while guardrail technology enforces security requirements and the platform operates entirely within the financial institution’s own environment without external connectivity.
- Data sovereignty is a core design principle: no financial data leaves the bank’s controlled environment, addressing the regulatory and security requirements that have made cloud-based AI adoption particularly slow in Japanese and Asian banking sectors.
What Fujitsu Announced
The announcement, published via JCN Newswire on July 28, 2026, describes a dedicated proprietary AI environment for financial institutions that combines three Fujitsu AI technologies: the Kozuchi Enterprise AI Factory as the foundational platform; the Takane LLM for financial language comprehension; and multi-AI agent technology for autonomous execution of banking workflows. The primary targets for the initial deployment are Japanese regional banks, which face particular pressure to modernise as consolidation, demographics, and competition from fintech companies stress their operational models. Fujitsu has framed the platform as extensible to international banks and financial institutions globally, and the architecture — built around data sovereignty and on-premises operation — is designed to be applicable in regulatory environments beyond Japan where financial data residency requirements constrain what AI infrastructure banks can deploy.
The Three Technology Layers
Kozuchi Enterprise AI Factory
The Fujitsu Kozuchi Enterprise AI Factory is the foundational platform on which the Finance AI Transformation Platform is built. Fujitsu launched the Kozuchi Enterprise AI Factory as a commercial product in January 2026 and began its official rollout in July 2026, having been in preliminary trial since February. The platform supports autonomous management of the entire AI lifecycle within an enterprise: model development, operation, incremental learning, and continuous improvement of both models and agents, without requiring cloud connectivity. It supports the Model Context Protocol (MCP) and inter-agent communication, allowing multiple AI agents to collaborate on complex, multi-step workflows. It also supports in-house fine-tuning and model quantisation, allowing enterprises to adapt base models to their specific terminology and workflow patterns without sending training data to external providers. For financial institutions, the on-premises architecture is the critical enabler: it is what allows a bank to use AI on its own customer and transaction data without that data leaving the institution’s controlled technical environment.
Takane: A Financial-Domain LLM
Takane is Fujitsu’s proprietary large language model, co-developed with Cohere Inc. Unlike general-purpose LLMs, Takane is trained on the specific corpus of materials that financial institutions work with: legal frameworks governing lending and deposits, banking customs and operational procedures, financial product terminology, regulatory compliance language, and the document formats that characterise Japanese and international financial workflows. The practical consequence of domain-specific training is that Takane does not need to be corrected on basic financial concepts, does not produce outputs that sound generically plausible but are technically wrong in financial context, and can generate first-draft documents in the specific formats and language conventions that banking operations require. General-purpose frontier models — GPT-5.6, Claude Fable, Gemini — are capable of discussing financial topics at a general level but lack the specific training on Japanese financial law, regional bank operational procedures, and FSA regulatory formats that Takane is designed to cover. The same domain-specific model principle that drove AT&T to build OTel 2.0 for telecom operations applies here: frontier models do not speak the specific language of any industry at the depth that production operations require.
Multi-AI Agents and Guardrail Technology
The multi-AI agent architecture is what allows the platform to handle complex, multi-step banking workflows rather than single-turn queries. A loan screening process involves multiple stages: initial eligibility check, document collection and verification, credit history analysis, regulatory compliance check, and decision documentation. Each stage requires different data, different analysis, and different output formats. Multi-AI agents can execute these stages autonomously and sequentially, passing outputs from one stage as inputs to the next, without requiring a human to manually coordinate the handoff. Guardrail technology enforces security requirements at each stage: preventing agents from accessing data they are not authorised to view, ensuring outputs meet compliance standards before being surfaced to human reviewers, and logging all agent actions for audit purposes. The combination of autonomous execution with security guardrails is specifically designed to address the concern that AI agents with broad access to sensitive financial data create unacceptable security exposure.
Why Data Sovereignty Is the Core Requirement
Financial institutions in Japan and across Asia operate under regulatory frameworks that impose strict requirements on how customer financial data is stored, processed, and accessed. Japan’s Act on the Protection of Personal Information, the Financial Services Agency’s cybersecurity guidelines, and sector-specific data governance requirements collectively make cloud-based AI processing of customer financial data legally and operationally complex. A bank that sends loan application data to an external AI provider’s API for processing must address questions about data residency, third-party access, audit trail requirements, and data retention that do not arise when the processing happens on the bank’s own infrastructure. The Uvance for Finance AI Transformation Platform’s on-premises, no-external-connectivity architecture sidesteps these questions entirely: the AI is physically located in the bank’s own data center or private cloud, processes data that never leaves the bank’s network perimeter, and generates outputs that are subject to the bank’s own access controls and audit systems. This architecture mirrors what Accenture built for the European Commission’s DG INTPA AI assistant, as covered in our earlier reporting on enterprise AI deployed inside compliance-constrained institutions — in both cases, the architectural requirement is the same: AI that operates inside the institution’s controlled environment rather than calling external services.
The Japanese Regional Banking Context
Fujitsu’s decision to target regional banks as the primary initial customer segment reflects the structural pressures facing those institutions in 2026. Japan has approximately 100 regional banks, many of which are facing simultaneous challenges: declining populations in their home prefectures reducing the customer base, competition from fintech companies and the Japan Post Bank, rising compliance costs from regulatory updates including new FSA digital asset guidelines, and the need to modernise legacy core banking systems that in some cases date back decades. AI-assisted loan screening and document processing offers regional banks the ability to handle the same volume of applications with fewer staff at a time when financial services employment is under pressure from both cost and demographics. The platform’s focus on use cases that are specific to regional banking — local business loan assessment, agricultural lending documentation, prefectural regulatory compliance — reflects Fujitsu’s understanding that a generic AI deployment is not sufficient for institutions whose competitive differentiation depends on their knowledge of local markets and customers that national banks lack.
The Broader Fujitsu Finance Strategy
The Uvance for Finance AI Transformation Platform is the latest addition to a financial services strategy that Fujitsu has been building through multiple partnerships and platform investments. In March 2025, Fujitsu partnered with FICO to bring FICO’s decisioning and fraud prevention platforms to Japan, providing the credit scoring and risk management analytics layer that complements the Uvance platform’s AI workflow capabilities. Earlier, Fujitsu signed a strategic collaboration agreement with Amazon Web Services in 2022 to accelerate digital transformation in the financial and retail industries. The pattern is a platform company strategy: Fujitsu positions the Kozuchi Enterprise AI Factory and Takane as the proprietary AI core, integrates third-party analytical and compliance tools through partnerships, and provides the system integration and implementation support that Japanese financial institutions need to go from a technology deployment to an operational change.
What Happens Next
Development of the Uvance for Finance AI Transformation Platform begins August 1, 2026. Fujitsu has not disclosed a commercial availability timeline, but the company’s prior pattern with Kozuchi-based platforms — preliminary trial registration in February, progressive feature rollout, official launch approximately five months later — suggests a commercial availability target in the range of late 2026 or early 2027. The regional bank customer segment that Fujitsu is targeting is well-defined and accessible through Fujitsu’s existing relationships with Japanese financial institutions: Fujitsu has maintained decades-long technology partnerships with major Japanese banks and regional financial institutions, giving the platform a direct customer pipeline. International financial institution expansion — referenced in the announcement as a longer-term target — will require the platform to be adapted for the specific regulatory frameworks of markets outside Japan. The architecture supports this adaptation because it is designed to operate within any financial institution’s own controlled environment, meaning regulatory compliance is a function of the bank’s own governance rather than Fujitsu’s external service. The same trend toward AI as critical enterprise infrastructure documented in our coverage of the AT&T 45 billion token AI Gateway and OTel 2.0 initiative applies here at the domain-specific model level: both Fujitsu and AT&T are building proprietary domain AI because frontier models do not speak the specific language of their industry at the precision their operations demand.
Why It Matters
The Fujitsu Uvance for Finance AI Transformation Platform matters because it addresses the specific combination of requirements that has prevented mainstream AI adoption in banking more than in any other major enterprise sector: domain-specific language capability, data sovereignty, multi-step autonomous workflow execution, and security guardrails — all in a single integrated platform that operates within the bank’s own controlled environment. Banking is simultaneously the sector with the most to gain from AI automation of high-volume, rules-driven workflows and the sector with the most regulatory and security constraints on how that AI can operate. A platform that genuinely satisfies both dimensions in a production-ready, on-premises architecture has a large and underserved market in front of it — not just in Japan, but across every jurisdiction where financial data residency requirements have blocked the cloud-based AI adoption path that other enterprise sectors have taken.
Sources
Fujitsu Limited official announcement via JCN Newswire / ScoopAsia, July 28, 2026. Fujitsu Kozuchi Enterprise AI Factory launch announcement via Nasdaq/JCN Newswire, January 26, 2026. Fujitsu and Cohere partnership background. MEXC News Fujitsu AI-Driven Software Development Platform coverage.
