Customer Service Automation · Enterprise Knowledge Search · Risk Management
The bank needed one enterprise large-model foundation that could serve marketing, risk control, document Q&A, coding, and customer-service scenarios instead of leaving AI capabilities fragmented across separate teams and tools.
Reported impact
The bank created a reusable generative-AI foundation that lowered development cost, improved knowledge retrieval and service quality, and gave multiple business lines a faster path from experimentation to production.
Baidu AI Cloud helped the bank build a bank-wide large-model stack on the Qianfan platform. At the infrastructure layer, the project established a heterogeneous model-service cluster to support both training and inference. At the platform layer, Qianfan unified data preparation, training, evaluation, tuning, and deployment so models could be governed across their full lifecycle. On top of that stack, the bank adopted enterprise-grade visual scenario debugging and document-data preprocessing to ground applications on internal financial and insurance content. Reusable AI-native services, including Wenxin Kuaima for developers, knowledge-retrieval middleware for document Q&A, and an intelligent customer-service assistant that summarizes live customer issues and recommends answers, were exposed to business teams so new use cases could launch on a shared governed platform rather than duplicated point solutions.
Internal business and insurance documents are preprocessed through the enterprise framework, then routed into Qianfan-managed large models running on the heterogeneous cluster. Grounded retrieval results and model outputs are served back into coding, knowledge-Q&A, and customer-service experiences, where agents and employees receive recommendations, summaries, and generated responses.
At the application layer, AI-native tools such as the code assistant and intelligent customer service can be used out of the box (original: 应用层，代码助手、智能客服等百度智能云AI原生应用可开箱即用).
Baidu AI Cloud helped the bank build a unified large-model technical system for the entire bank (original: 百度智能云助力该股份制商业银行建设全行统一的大模型技术体系).
Knowledge-retrieval coverage accuracy can reach more than 85%, and customer-service recommendation accuracy is expected to rise from 30% to above 60% while average talk time drops by 15 seconds (original: 检索覆盖准确率可达85%+；预计可将知识推荐的准确率从传统NLP的30%提升到60%以上，降低平均通话时长15秒).
The development framework provides enterprise-grade visual scenario debugging and enterprise document data preprocessing (original: 为大模型应用提供企业级的场景可视化调试、企业文档数据预处理能力等支持).
The project built a heterogeneous service cluster at the AI infrastructure layer and unified training, evaluation, and optimization at the platform layer (original: 在AI基础设施层，建设大模型异构服务集群; 平台层，对大模型全生命周期的数据训练、评估、优化等流程进行统一管理).
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 · Sep 17, 2026 Home Credit Bank Home Credit Bank needed a safer enterprise AI environment that non-technical teams could use to build banking agents quickly without creating data-leak risk or relying on a long custom-development cycle for each new use… Achieved 100% migration to a secure AI environment Agent Designer · Gemini Enterprise 
OpenAI Sample Jun 11, 2026 BBVA BBVA needed a secure, governed way to embed AI across a global bank so customer experiences, risk analysis, operations, and employee productivity could all improve without sacrificing regulatory discipline. 20,000+ GPTs created across the organization ChatGPT Enterprise · Client Experience Assistant GPT · Credit Analysis Pro GPT 
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 
