Coercion is All You Need? The Introduction of Automation Technology to the Tech Workplace

• In Event: AI and the Value of Work

Abstract

How are AI coding tools being introduced to software engineering workplaces, and how are workers responding? In this interview study of software engineers at tech companies in the United States, I investigate the labor process of these workers and how it is changing after the introduction of LLM-based productivity tools for coding and writing. I find that, in the initial stages, this is not the deskilling of the 20th century. The knowledge of the labor process currently still rests with the workers themselves, and so the responsibility for discovering how to realize returns on AI tool investment is passed on to rank and file workers. This leads executives to make commands about use of the tools that are often mismatched to workers’ understandings of what effective uses of the technology can be. These workers have high expectations of autonomy, and their labor market position is based solely on their specialized expertise. Therefore, commands and coercion from management to use LLM-based tools undermines their professional self-identity and threatens their position on the labor market. I find that workers handle these competing needs by integrating LLMs into their professional identity in nonthreatening ways. “Adopters” expand their professional jurisdiction to include LLMs. “Jockeys” attempt to enhance their power in the internal labor market of the firm by choosing to administer these mandatory tools to other workers. “Peripheral users” do local decoupling, conforming to the command to use the tools but only on tasks they can exclude from the technical core of their expertise. “Artisans,” who see code as craft, reject integration entirely, seeing LLMs as threats to their specialized expertise. Finally, “Ideologues” resist integrating LLM use into their professional jurisdiction because they chose the profession specifically to avoid work with negative social externalities.

Author

• Emily Mazo, Columbia University