Across South Africa and the broader African continent, a paradox is playing out: highly educated professionals, PhD holders, researchers, and domain experts are being hired by AI companies to teach machines the very judgment they spent years developing. The pay is modest by global standards but attractive locally, creating a difficult moral calculus for workers who know the systems they are building could eventually replace them.
James Maisiri, who holds a PhD in industrial sociology from the University of Johannesburg, describes the dilemma plainly. When the opportunity to earn a livelihood collides with your principle of helping a technology that can compete with you, what do you do? The answer, for many, is shaped by socioeconomic reality. Unemployment is high, wages are low, and the money is hard to refuse.
The work being outsourced has shifted from simple data labeling to something far more nuanced. AI companies now need experts to transfer professional judgment, including legal reasoning, clinical assessment, and pedagogical knowledge, into training data. These are not annotation tasks; they are knowledge extraction exercises that require deep domain expertise.
The trend raises uncomfortable questions about the global AI economy. Africa has among the youngest populations and a growing pool of educated professionals facing limited opportunities. For AI companies, this creates a cost-effective source of expert training data. For the workers, it means participating in building systems that may eliminate the very expertise they are being paid to share.
