Terasios Bundi spent twelve years writing other people's coursework from Kenya. More than 2,500 papers, and by the end he was charging $40 to $70 for a single text. He is 34. ChatGPT took the demand and the rates together, and it did not take them from one freelancer — academic ghostwriting was one of several online trades Kenya had spent a decade cultivating, and it was the one that machine-generated text replaced most directly.
The rest of the country's online work contracted as well: audio transcription, data labeling, and content moderation for Meta.
None of this grew by accident. Since 2016 the Kenyan government deliberately built up online gig work as an economic policy, and companies such as Samasource recruited Kenyans to label the data that AI systems are trained on. Around 80% of jobs in the country sit in the informal sector, which is the number that turns a market shift into a household emergency: informal work comes with no severance, no notice period, and no line in anyone's unemployment statistics.
What remains, according to The New York Times, are the "humanizers" — people who rework AI-generated text so it passes plagiarism checks.
That job is the same trade after compression. The writer who once produced the document now edits the machine's output so it can be submitted as a person's. The skill required is narrower, the unit of work is smaller, and the customer is buying laundering rather than writing. It reads less like a new profession than like the last billable step left over after the rest of the task was automated, and there is no obvious reason it survives the next model that writes less detectably.
The sequence is worth sitting with. Kenya was paid to prepare training data for AI systems, then lost work to AI systems. The source does not claim one caused the other, and I am not claiming it either — but a country that positioned itself as the labor layer of someone else's automation stack had, by construction, no ownership of what the stack would eventually do.
Notably absent from the reporting: any count of how many people did this work, or what the market was worth. The collapse of Nairobi's essay industry is easy to describe through one man with 2,500 papers behind him and impossible to size. That asymmetry is not incidental. Informal work is invisible on the way up, which is why it can also disappear without registering anywhere as a loss.
Mark Graham, a professor at Oxford University, expects shocks like this worldwide. Bundi's own list of what comes next is bankers, accountants, engineers, architects — and eventually, he thinks, everyone.
His forecast carries weight not because he can model labor markets but because he is describing something behind him rather than ahead of him. The difference between his trade and the ones he names is not exposure to the technology. It is that when it happens to accountants and architects, someone will be counting.