With the continuous advancement of artificial intelligence technology, the ability to automatically convert natural language instructions into executable code has become a cutting-edge research focus. To apply NL2Code in the field of backend business code generation, NL2SQL (natural language to structured query language) and template engine-based code generation techniques have been adopted. By conducting coding experiments using the Java language, it is possible to quickly generate backend business code from natural language instructions. Therefore, the new method of generating backend business code from natural language proposed in this paper enables more efficient and accurate automatic code generation. This approach not only leverages the strengths of AI in natural language processing but also addresses the practical needs of software development, bridging the gap between non-technical user input and technical code output. The integration of NL2SQL and template engines ensures that the generated code is both functional and maintainable, marking a significant step forward in automating the software development process.