Human-AI Work Redistribution Across the Software Development Life Cycle: A Socio-Technical Study of Indonesian Developers

Abstract

AI Coding Assistants are becoming increasingly integrated into software development, changing not only how developers perform their tasks but also how humans and AI share responsibilities. However, how this redistribution of work occurs across the Software Development Life Cycle (SDLC) remains insufficiently understood. This study investigates how responsibilities are distributed between developers and AI across SDLC activities, how these changes affect software engineering workflows and developer roles, and which socio-technical factors influence this process. Semi-structured interviews with seven software professionals in Indonesia were analyzed using reflexive thematic analysis. The results identify three patterns of human-AI work redistribution: AI-delegated, AI-augmented, and human-controlled work. AI contributes more substantially to structured, repetitive, and relatively verifiable activities, particularly code generation, debugging, testing, and documentation. In contrast, requirements analysis, architectural decisions, security assessment, and release approval remain largely under human control because they require contextual understanding, professional judgment, and accountability. The findings also indicate a shift in developer responsibilities. This redistribution depends on developer experience, AI literacy, trust, model capability, task complexity, security requirements, team practices, and organizational policies. Drawing on socio-technical systems theory and activity theory, the study develops an empirically grounded Human-AI Work Redistribution Model that conceptualizes AI-assisted software engineering as a selective redistribution of responsibilities rather than uniform automation across the SDLC.