Cheap Code, Costly Judgment: A Case Study on Governable Agentic Software Engineering

2026-07-01

AI Summary

This study addresses the challenge of ensuring auditability, correctability, and maintainability in software development within AI-driven, high-velocity code generation environments. Drawing on a 12-week field case study that integrates AI coding agents, ethnographic field notes, and over 1.5 million lines of code artifacts—including 420 KLOC of production code—the work proposes a mid-range theory of “governance translation.” This model elucidates how engineers transform structured failures into enduring governance mechanisms through professional judgment, thereby transcending traditional governance paradigms reliant on predefined obligations. It highlights the pivotal role of engineering judgment in balancing the speed of AI-assisted development with system controllability, offering a novel, empirically grounded theoretical framework and actionable pathways for software engineering in the age of AI.

Abstract

Generative AI is shifting software engineering from a practice organized around scarce implementation effort toward one organized around abundant, low-cost code production. This shift changes the central engineering problem: not whether AI can generate useful code, but how engineers organize architectures, tools, evidence, and feedback loops so that AI-mediated development remains inspectable, correctable, and maintainable. We study this problem through a first-person case study: a 12-week development effort in which a single expert software engineer used frontier AI coding agents to build a document accessibility remediation system. The empirical record comprises 88 contemporaneous field notes, 420 KLOC of production code, and 1.16 MLOC of tests, lints, supporting documentation, and agent tooling. From this record, we develop a candidate middle-range theory of governance conversion, expressed as a process model explaining how high-velocity agentic implementation becomes governable. The model explains how agentic implementation velocity surfaces recurring structural failure classes, and how engineering judgment sustains velocity by converting those failures into durable governance mechanisms. In contrast to existing governance models that derive controls from known obligations, governance conversion explains how controls are discovered from failures that become visible only during agentic work. We use our model to make testable predictions and to describe implications for software engineering research and practice.

Problem

• agentic software engineering
• governable AI
• software maintainability
• engineering judgment
• AI-mediated development

Innovation

• governance conversion
• agentic software engineering
• AI-mediated development
• engineering judgment
• failure-driven governance