TL;DR
- James Maisiri describes being offered 600 rand, about $37, an hour to train an AI system in teaching and assessment.
- The work would draw on professional judgement: how an experienced teacher weighs context and explains a decision.
- His account helps professionals consider the contribution, intended use and contract terms alongside the rate on offer.
- He stopped pursuing the role after an AI interview and remains uncertain about his reasons.
In plain English
Expert training data includes examples, corrections and assessments supplied by people with relevant professional knowledge.
Professional judgement is the ability to weigh the details of a situation and explain an appropriate decision, such as distinguishing understanding from memorisation in an essay.
Automated recruitment uses software for parts of hiring, including interviews or assessments. Maisiri reports an AI interview and automated feedback in his application process.
James Maisiri expected to teach after finishing his PhD. His first offer came from a recruiter looking for someone to train an AI system to design assessments, teach undergraduates and mark essays. In his account for Rest of World, he explains what that request meant to someone who had spent years learning to do those things himself.
The role offered 600 rand, about $37, an hour. Maisiri sets that beside South Africa’s minimum wage of 30.23 rand and youth unemployment of 47.4% in the second quarter of 2026. Those figures, as reported in his essay, explain part of the context in which he considered the offer: a qualified professional seeking work, with a paid opportunity to contribute his expertise to AI development. He extends the argument to Africa more broadly: a young, educated workforce facing unemployment and low wages can make professional expertise attractive to AI companies seeking lower costs. In his account, limited local adoption of AI sits alongside the opportunity to supply the labour used to develop it.
The contribution he describes goes beyond supplying course material. Teaching had helped him judge which concepts matter, distinguish memorisation from understanding and explain why an essay deserves 75% rather than 60%. Those decisions depend on context and standards developed through practice. Showing how he reaches them would provide examples for a system being trained to perform similar work.
That is the central idea of the essay. AI-training work can ask a professional to make their judgement explicit: which details they notice, how they weigh them and why they reach one conclusion rather than another. An answer and an explanation of how it was assessed provide different kinds of material to learn from.
Maisiri places his experience alongside other forms of training work. He describes workers in India recording physical tasks for robot training, and names platforms such as Outlier, Mercor and Surge as recruiters of professional expertise. The examples span different work and pay, but help explain the range of human skill being gathered as training material.
For him, that raises a question about professional identity as well as employment. The expertise on offer had taken years to develop and was the basis of the academic work he hoped to do. Contributing it to AI training offered a way to earn money while also raising questions about how the resulting systems might affect that work in future.
His recruitment experience adds another part of the picture. Maisiri says an AI interviewed him for 45 minutes, sent feedback on his strengths and weaknesses, and suggested he retake part of the assessment. He did not retake it and stopped pursuing the job. He describes no direct exchange with a human recruiter during that process.
He leaves his reasons for walking away unresolved. “I still do not know why I walked away from the AI training job,” he writes. The uncertainty is part of the account: he is working through what it means to contribute the judgement associated with his profession to a system that may eventually perform some of that work.
For professionals considering similar offers, the essay gives the decision some useful detail. What will you be asked to demonstrate? How will your examples be used? What does the contract allow the buyer to retain or reuse? Those questions sit alongside the hourly rate and the alternatives available to you.
For teams commissioning expert training data, it also explains why the professional’s contribution matters. If the work depends on context, the examples need to preserve enough of that context to make the assessment useful. Collecting a grade alone gives a different starting point from collecting the reasons an experienced teacher gave that grade.
Maisiri’s essay describes one person’s experience, without establishing what a resulting model will be able to do. Its value is the account of expertise being translated into training work and the choices that creates for the person providing it. That is a concrete part of the AI economy: people deciding how to earn from skills they developed for another kind of work, and what they are willing to pass on.