Two-Pronged Approach: JRCB Advances AI Innovation in Banking with Huawei Using Models and Agents
Przedsiębiorstwa
While many organizations use large AI models primarily to improve efficiency, JRCB has positioned AI models as a key driving force of future growth from the outset. The bank has developed a roadmap built around 1 vision, 3 goals, and 4 strategies, with a focus on improving efficiency, transforming processes, and optimizing organizational structures. This will help the bank take a leading position amid the fintech waves.
A key feature of this roadmap is its phased and pragmatic approach to AI adoption, which prioritizes practical applications before expanding into more complex banking operations. Guided by the principles of technology leadership, service orientation, internal prioritization, and independent innovation, JRCB initially focused on internal applications, including intelligent OA, R&D, and operations, to reduce employee workloads and establish a solid technical foundation. As its AI capabilities mature, the bank plans to expand into higher-value areas such as intelligent risk management and marketing. This gradual approach enables JRCB to achieve innovation while maintaining service stability. To date, the bank has successfully incubated and rolled out more than 100 use cases, including internal knowledge Q&A, intelligent customer service, and credit contract review.
As large AI models move beyond general‑purpose Q&A into specialized applications, financial institutions face increasingly complex challenges. Core banking services, such as credit approval and contract review, involve intricate workflows, specialized knowledge, and stringent risk management requirements that general-purpose models alone may struggle to meet. The challenge is not simply enabling AI to understand requests, but equipping it to execute complex tasks competently.
To address these challenges, JRCB partnered with Huawei to jointly develop financial agents capable of understanding requests, planning workflows, and executing complex tasks. This collaboration has resulted in two key applications: a credit report review agent and a loan contract review agent. Rather than functioning as mere Q&A chatbots, these agents are integrated into the bank's operational systems to support critical review processes. These capabilities are supported by Huawei's Atlas AI foundation, built on the DCS AI solution. The foundation brings together Atlas series AI servers, OceanStor Dorado all-flash storage, CloudEngine network infrastructure, FusionCompute virtualization, and the ModelEngine end-to-end AI toolchain. The advantages of Huawei product portfolio and a full-stack solution enable implementation of three core capabilities of the DCS AI solution—data enablement, model enablement, and application enablement, helping the customer build an efficient, stable, and scalable AI factory that unlocks data value by managing the entire AI life cycle.
Building on this foundation, JRCB combines its financial expertise with large AI model capabilities. Take the loan contract review agent as an example. When a new loan application reaches the review stage, the agent automatically reads and analyzes the contract, evaluates its clauses at multiple criteria against a compliance rule library and a risk model, and identifies potential risks.
The collaboration has already delivered measurable results. Both agents have been deployed in the bank's production system. The loan contract review agent is called by more than 50 times per day, achieving an accuracy rate of 80% and reducing review time for employees by approximately 30%. Meanwhile, the credit report review agent operates reliably, helping improve the efficiency and consistency of credit reviews. These results demonstrate how AI is evolving from a supporting tool into an integral part of banking operations, allowing employees to devote more time to higher-value activities such as risk assessment and customer service.
Beyond developing AI agents with Huawei, JRCB is also working to strengthen its sustainable AI capabilities. The bank deeply recognizes the power of a two pronged strategy. It combines external general purpose large AI models with a bank specific model trained on its private data to generate strong synergies for banking operations.
To achieve this, JRCB began fine-tuning its own dedicated large AI model using extensive policies and regulations, banking knowledge bases, and AI feedback data. Through targeted training on open-source large AI models and multiple rounds of iterations, the dedicated model has achieved an 84% accuracy rate in financial-domain responses, matching or outperforming larger general-purpose models in certain scenarios while maintaining strong general knowledge capabilities.
This effort has enabled JRCB to develop a dedicated model with extensive banking knowledge, supported by a comprehensive toolchain for data engineering, model training, evaluation, and management. As a result, the bank is moving beyond simply adopting AI technologies to continuously refining and optimizing its own capabilities, laying a stronger foundation for future innovation.
Looking ahead, JRCB plans to develop a superior service portal powered by multi-agent collaboration. Under this approach, a central agent will interpret user intent and coordinate tasks, while a network of specialized agents handles functions such as tool invocation and data analysis. By connecting capabilities across different applications, this architecture aims to deliver a one-stop, seamless, and intelligent service experience for both customers and employees.
From strategic planning and joint AI agent development with Huawei to building its own specialized model, JRCB's AI journey offers valuable insights for other regional banks pursuing digital and intelligent transformation. Its experience demonstrates how a clear strategy, a pragmatic approach, and collaboration with a strong technology partner can help financial institutions translate AI innovation into practical results and deliver greater value to customers.