In early 2025, with the rise of vibe coding, it seemed that AI agents would turn software development into a near-direct transformation of prompts into working programs. Reality turned out to be more complex. While code generation can be delegated, the complexity of development does not disappear—it shifts. The challenge moves to defining tasks, structuring context, managing constraints, and maintaining control over how a product evolves.
This book addresses that shift.
This book introduces Agent-Driven Software Management (ADSM), an original methodology developed by the author. It provides a practical approach to delegating development to AI agents while maintaining control, preserving product integrity, and ensuring that results remain suitable for further development.
ADSM is based on real-world experience. In the author's work, code generation for a new version of the Tequila Framework platform was fully delegated to agents, while the primary effort moved to documentation, context design, and execution control. This book distills that experience into a structured approach.
If you are already using AI agents in development and facing issues with instability, unpredictability, or loss of control, this book offers a way to turn that experience into a disciplined engineering process.
