The Productivity Paradox of AI-Powered Development

Abstract

Over the decades, software engineering has suffered from trends seeking to manage technical and non-technical issues and to improve the software productivity. Generative AI has emerged as a powerful tool, promising to automate tasks and increase efficiency. This study explores, in an industrial setting, how developers’ perceived productivity when using AI-based code assistants relates to key DevEX dimensions. Our results reveal a paradox: AI tools can both enhance and hinder productivity, depending on interaction patterns, cognitive load, and contextual factors.