Marat, a developer with 15 years of experience in the SKIF inventory management system, discusses the feasibility of using AI to build custom ERP solutions. While AI can assist with simple tasks like adding columns to reports, the author notes that core accounting logic requires human decision-making regarding business rules, such as handling negative inventory or tax rounding.
The author highlights that AI often struggles with the specifics of Russian accounting standards, including VAT, electronic document management, and labeling. While AI can generate code that appears correct, it frequently produces errors in complex integrations, making manual development more reliable for critical business infrastructure.
Building a functional ERP system involves thousands of lines of code and significant maintenance. The author emphasizes that software only matures through real-world usage, where user behavior often reveals edge cases that developers cannot anticipate. Furthermore, relying on AI-generated code lacks the accountability provided by professional software vendors.
The article concludes that businesses should prioritize a stable, mature foundation for their operations. While users can leverage AI for minor customizations or 'vibe-coding' on top of an existing system, the responsibility for core functionality and maintenance is best left to experienced developers who can provide support and practical, experience-based solutions.
