Experimental AI at Work: Labour, Technological Uncertainty, and the Workplace as Laboratory

AI systems are increasingly deployed across workplaces while their capabilities, risks and consequences remain under experimentation and evaluation.

This article develops the concept of experimental AI to examine how the boundaries between technological testing and workplace deployment are becoming increasingly blurred. Extending labour process theory, we argue that AI introduces a distinctive form of experimental indeterminacy. Alongside the enduring indeterminacy of labour, uncertainty persists about technological systems themselves. We examine how this uncertainty is organised through private industry and infrastructures of business law, including corporate testing, technical standards, conformity assessment and regulatory experimentation.

As workplaces become sites through which technological uncertainty is experimentally resolved, workers may contribute simultaneously to production and to producing knowledge about technologies that reorganise their labour. Bringing work and employment into dialogue with science and technology studies and legal scholarship, we ask who controls experimentation, who benefits from the knowledge it produces, who bears its risks and uncertainties, and what rights workers possess when workplaces become experimental spaces.