Kingdee Develops Generative AI Features in Two Months on Amazon Bedrock, Enhancing ERP User Experience

Kingdee provides enterprise management and cloud-based ERP solutions, supporting millions of businesses worldwide. Despite the powerful capabilities of ERP systems, they often present usability challenges, especially when users need to retrieve complex data or complete multi-step tasks, like processing expense claims or accessing detailed financial metrics. Traditional ERP interfaces require navigating multiple screens or reports, which can slow operations and create friction for users. Kingdee recognized an opportunity to simplify ERP interactions and align with user expectations for speed and efficiency.

Having collaborated with AWS since 2015 on its enterprise SaaS solutions, including Kingdee Constellation, Kingdee Galaxy, Kingdee Stellar, and EAS Cloud, Kingdee turned to AWS again to drive innovation in ERP accessibility using generative AI. As part of its strategic "All in AI" initiative, Kingdee aimed to transform ERP functionality with generative AI-driven interfaces, specifically focusing on intelligent ticketing and intelligent metrics analysis to enhance the ERP user experience.

Fang Shouyu, intelligence finance product manager at Kingdee, says, “After evaluating multiple products, we found AWS generative AI services to be both robust and user-friendly, offering mature solutions backed by exceptional support. AWS gave us the confidence to drive innovation in ERP user experience through generative AI."

To meet the evolving demands of ERP systems, Kingdee developed two generative AI solutions powered by Amazon Bedrock: an intelligent ticketing system and a metrics analysis tool. Amazon Bedrock provided a selection of foundational models (FMs), customization options, and application integration features, giving Kingdee access to mainstream models for its global operations while maintaining enterprise-level security. Kingdee also implemented Amazon Bedrock Agents to develop generative AI applications capable of executing complex tasks across company systems and data sources. By integrating with AWS serverless services like AWS Lambda and Kingdee’s various databases, Amazon Bedrock Agents helped Kingdee rapidly implement generative AI in ERP software applications, utilizing Kingdee’s APIs and knowledge base to create dedicated agent applications​.

In January 2024, following a detailed evaluation, Kingdee completed the proof of concept (POC) for its intelligent ticketing system within a week, quickly rolling it out to international clients who provided positive feedback. Fang says, “Amazon Bedrock Agents stood out for its maturity and ease of use. After testing AWS’s tools, we quickly recognized its potential. Even as new frameworks emerged in 2024, Amazon Bedrock Agents remained unmatched in usability, offering timely monitoring, quick debugging, and a smooth development process that helped us deploy generative AI applications efficiently on a global scale.”

Additionally, Kingdee’s intelligent metrics analysis tool offers enterprise managers direct access to metrics and accurate results through natural language queries. Using the Cohere embed-multilingual model and Amazon OpenSearch Service, the tool allows managers to retrieve data instantly, eliminating the need to navigate complex BI reports. Acting as a 24/7 virtual assistant, it delivers timely insights, simplifying metrics comparisons and financial analysis. Kingdee also employs Anthropic’s Claude 3 Sonnet model in Amazon Bedrock for natural language understanding, task decomposition, and parameter extraction. The integration of Text2SQL with Amazon Bedrock Agents further enhances query flexibility by converting natural language requests into SQL queries, ensuring fast and precise access to ERP data​.

Kingdee’s intelligent ticketing system has transformed ERP interactions by allowing users to complete tasks through natural language commands instead of navigating a complex interface. Common requests, such as submitting expense reports or applying for loans, can now be completed with simple inputs, significantly enhancing user experience and productivity. Fang explains, “With the intelligent ticketing system built on AWS, users only need to inform the system of their status and the task they want to complete. The system then makes inferences based on the relevant rules, provides corresponding guidance, and feeds back to the user.”

In addition, Kingdee’s intelligent metrics analysis tool empowers enterprise managers with direct access to financial and operational insights at high accuracy rates. For straightforward queries, the system achieves over 90 percent response accuracy in the first round, while complex queries, such as trend calculations, reach 70 to 80 percent accuracy initially, achieving 100 percent accuracy in follow-up responses. This improvement has streamlined data access, supporting more efficient decision-making and optimizing ERP usability for Kingdee’s clients.

Comprehensive support and training from AWS facilitated the swift deployment of these generative AI applications. Fang says, “The AWS technical team brings a wealth of experience and expertise in generative AI applications. Their support throughout our product deployment and in tackling technical challenges has been invaluable, and we’re very pleased with their professionalism.”

Kingdee successfully implemented generative AI features within two months on AWS, rapidly addressing market demand and establishing a solid foundation for future AI-driven advancements. As Kingdee continues its “All in AI” initiative, the company expects to drive further innovation in ERP applications. Fang says, “By working with AWS, we’ve been able to push the boundaries of ERP usability through generative AI. AWS’s commitment to supporting our innovation aligns closely with our AI strategy, and we’re excited to keep expanding our AI capabilities alongside them.”