On October 29, at the 2025 Financial Street Forum, the case "Agentar Knowledge Engineering KBase Promotes Intelligent Upgrading of Financial Business," jointly created by Ant Digital and Ningbo Bank, was successfully selected as a "2025 International Standard Financial Application Excellence Case." This solution demonstrated outstanding performance in the application of knowledge engineering technology in the financial field. With core advantages such as high security, high accuracy, and strong interpretability, it received unanimous recognition from the evaluation experts, setting a new benchmark for the intelligent upgrading of the industry.

    Currently, financial institutions generally face the challenge of "knowledge silos": key information such as user characteristics data, product rules, and market research is scattered across different systems, leading to issues such as inaccurate knowledge retrieval, low service efficiency, and poor professional consultation experience. How to transform dormant knowledge assets into core competitiveness has become an important issue in the digital transformation of the financial industry.

    To address these pain points, Ningbo Bank and Ant Digital collaborated, leveraging the Agentar Knowledge Engineering Platform (KBase), to build an intelligent decision-making system that integrates "knowledge processing and management platform + logical reasoning engine + intelligent application scenarios," providing a solid technical foundation for the bank's intelligent upgrading.

    The platform achieves full lifecycle management of multi-source heterogeneous data inside and outside the bank, with core capabilities including intelligent Q&A, knowledge processing, multi-channel recall, and knowledge management. Its technological breakthroughs mainly lie in the knowledge-enhanced generation engine, which uses a collaborative mechanism of three operators: "planning - retrieval - reasoning," innovatively enabling bidirectional indexing between knowledge graphs and original texts, significantly improving the quality of knowledge. The system has achieved a key leap from "fuzzy matching" to "precise reasoning," increasing the depth of reasoning from one hop to 3-5 hops, allowing AI not only to "understand" financial expertise but also to possess human-like logical reasoning capabilities.

    Currently, this solution has been widely implemented in multiple internal scenarios at Ningbo Bank, such as market analysis, product interpretation, conversation training, and report writing. Evaluation results show that the accuracy of complex Q&A has increased from 68% to 91%, and response speed has reached the hundred-millisecond level. Among them, the accuracy of content recommendations has increased by 35%, and the recall rate has increased by 40%, achieving significant improvements in business efficiency.

    Notably, the Agentar Knowledge Engineering Platform has strong interpretability, capable of clearly displaying the reasoning path and data sources of answers, providing an important guarantee for the safe and compliant deployment of generative AI in the highly regulated financial environment.

    In the future, both parties will continue to deepen their cooperation, expand the relevant technologies to more extensive financial business scenarios, and actively participate in the construction of industry standards, promoting the standardized and scaled application of knowledge engineering and large model technologies in the financial field.