AI wrote all the code. Why did I have to build the PoC architecture and deployment myself?

Modern AI tools allow for the generation of software code in a matter of hours, yet transforming that code into a functional Proof of Concept (PoC) requires deep engineering expertise. In this article, the author analyzes the typical architecture of an AI project using a personal pet project as an example, emphasizing that writing code is only a small part of the process. The main challenges lie in architectural design, infrastructure setup, security, and organizational tasks that AI cannot yet effectively solve on its own. The author concludes that the role of the engineer is transforming from writing lines of code to managing complex systems and integrating AI solutions into real-world environments. The article aims to answer why human oversight remains a critical factor in software development, even when utilizing advanced generative models.