From Code Generation to Code Orchestration: Evaluating the Impact of Agentic AI on Software Development Productivity, Quality and Security

Abstract

Software engineering is in the middle of a quiet but far-reaching handover. For most of the last decade, artificial intelligence in the developer's tool chain meant auto complete: a model that finished a line, or occasionally a function, while a human wrote and reviewed everything around it. That arrangement has started to break down. Coding agents such as Claude Code, OpenAI's Codex CLI, Google's Jules, Devin, and Open Hands can now read an entire repository, plan a multifile change, run the test suite, and iterate on failures with little moment-to-moment supervision. The developer's job is shifting from typing code to directing agents that type code a change often summarized as a move from code generation to code orchestration. This paper reviews recent empirical literature to ask what that shift is actually producing.