Session Abstract

AI coding agents can generate features in minutes, write tests, refactor modules and suggest architectural changes — and they will do all of it with the same confidence regardless of whether your project has any structure or none at all. That is the actual problem. In this session, we will explore what happens when the same AI agent works inside a disciplined engineering workflow versus without one, using a live Next.js application built entirely with GitHub Copilot inside Visual Studio Code. We will compare two versions of the same feature: one built without guardrails, one built with them. The difference is not subtle. The structured version introduced 60% fewer issues overall, 86% fewer critical bugs, and prevented 12 specific defects before a single line of review was written - not because the AI suddenly became smarter, but because the context it worked with was better. The session is not about AI replacing developers. It is about a concrete, observable truth: the quality of AI-generated code is a direct reflection of your architecture guidance, conventions, and repository context. Engineering discipline is no longer just good practice. It is the interface between your team and your AI tooling. You will leave with a practical understanding of how to structure a codebase so that AI agents become a reliable part of your workflow rather than an unpredictable accelerant.

Jelena Ruzic

Jelena Ruzic is MA. Eng(IT) and Software Engineer. She is regional speaker and Microsoft Most Valuable Professional (MVP) for Developer Technologies category. Very active in IT community as a Co-leader of BLbit User Group in Banja Luka and Co-organizer of IT conference INIT.