Customer Service Automation · Document Processing · Risk Management
Postal Savings Bank of China needed a scalable AI infrastructure and unified model platform so it could automate customer service, trading, document review, and risk workflows without creating disconnected AI pilots across the bank.
Reported impact
PSBC built a reusable AI platform that improved customer interactions, accelerated institutional trading workflows, and automated large parts of loan-review processing while improving risk and compliance controls.
Postal Savings Bank of China used Huawei Cloud Stack to build an evolvable AI computing and data center plus the 'Youzhi' large model family for bank-wide intelligence. The platform follows a '1+2+3' design: one data standard, two data-set systems for financial and non-financial data, and three core platforms for compute management, development tools, and public services. This base supports model training and AI application development across more than 230 front-, middle-, and back-office scenarios. At the serving layer, PSBC combines large and small models for customer service, uses large models with RPA to automate trading and middle-office workflows, and fuses expert knowledge with multimodal data for risk analysis. Applications include mobile-banking smart customer service, a trading robot for institutional RFQ conversations, and an intelligent loan-review assistant that classifies images, validates materials, extracts key fields, and produces reports.
Financial and non-financial data is standardized and fed into the bank-wide AI platforms, where the Youzhi model family and related automation tools are trained and orchestrated. Outputs from customer-service, trading, and loan-review models are then delivered into front-, middle-, and back-office banking workflows.
"PSBC and Huawei Cloud Stack built an evolvable AI computing data center and used the 'Youzhi' large model to support business innovation from infrastructure up to intelligent applications. (original: 中国邮政储蓄银行携手华为云Stack，构建可持续演进的高效智算数据中心，以“邮智”大模型赋能业务场景创新实践，实现从算力基础设施到业务智能应用的双向突破。)"
"The bank built one data standard, two dataset systems, and three core platforms to support model training and intelligent application development across more than 230 scenarios. (original: 基于“1+2+3”的顶层设计，系统化构建了1套数据规范，金融、非金融2大数据集体系，以及算力管理、开发工具、公共服务3大核心平台... 围绕经营营销、对客服务、风险管理、智慧运营以及智能办公等重点领域覆盖前、中、后台230多个智能应用场景。)"
"The intelligent loan-review assistant reaches 96% image-classification accuracy, 92% key-information extraction accuracy, and can generate a review report within 20 minutes. (original: 智能审贷助手的影像分类准确率可达96%，关键信息提取准确率可达92%，且能在20分钟内生成审查报告。)"
"The trading robot has already handled inquiries from about 300 institutions, processed more than 1.5 trillion yuan in RFQs, and reduced average inquiry time to 22 seconds, about 94% faster than manual work. (original: 交易机器人至今已受理来自约300家机构的询价，询价总量超1.5万亿元，单笔询价交易平均耗时仅22秒，较人工询价时间缩短约94%。)"
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Google Cloud Sample Archived · Jan 16, 2026 Hong Leong Bank Hong Leong Bank needed to replace slow manual banking operations and limited chatbot performance with a secure AI ecosystem that could support dynamic customer support, document-heavy trade operations, sanction… 30% efficiency gain Agent Development Kit (ADK) · Cloud SQL · Compute Engine 
In-house Sample May 8, 2024 Mapfre Mapfre wanted to improve customer experience and streamline insurance operations such as claims handling, risk assessment, underwriting support, document management, internal knowledge access, and digital content… 75 use cases under study for the application of generative AI ControlExpert · Cyberwrite · Holistic AI In-house Sample Apr 9, 2024 ClearBank ClearBank needed to deploy multiple generative AI use cases for banking API workflows in a regulated financial environment while safeguarding security, robustness, transparency, bias control, and regulatory alignment. Large Language Models (LLMs) · evaluation pipeline · model experiment tracking system 
