AI ended a Nairobi industry in two years. The industry was cheating

Teresios Bundi is 34. Over twelve years he wrote more than 2,500 university papers for students in the United States and Britain, charging $40 to $70 each, at least five times what public-health work paid him in Kenya. He left a farming village for Nairobi in 2011 to become the first in his family at university, and a friend pointed him at the job. The New York Times found him this week, and his trade is gone.

ChatGPT arrived in November 2022. Within two years the orders and the prices had collapsed.

“A.I. is coming for bankers, for accountants, it's coming for engineers, and A.I. will come for architects. It's coming for everybody.”

— Teresios Bundi, academic writer, The New York Times, 5 September 2026

Quote source: Teresios Bundi, quoted by The New York Times, 5 September 2026

Kenya essay writing paid better than public health

Researchers counted at least 40,000 people working in Nairobi alone at the peak, writing medicine, computer science and engineering assignments for students who sometimes handed over their own university logins. Writers who are still working have watched monthly pay fall from $900 to $1,200 down to $500 to $800. One operator who ran a hundred writers has shut down. The Decoder summarised the reporting on Monday.

The product these 40,000 people sold was academic fraud. Their customers were students paying to avoid the work they had enrolled to do, and those students still pay. The survivors now work as humanizers, rewriting machine text until it slips past a plagiarism checker.

Same buyers, same dishonesty, tens of thousands fewer people doing it for roughly half the money.

The legitimate work went too

Kenya lost its transcription work, its data annotation work and its content moderation contracts for Meta over the same stretch. None of that was fraud. Those were the legitimate jobs, and they went the same way as the essays.

Samasource and companies like it built their Kenyan operations labelling the training data that machine learning systems need, and the government had promoted online gig work as policy since 2016. Kenyan workers helped teach the models that then took their work.

About 80% of jobs in Kenya sit in the informal economy, which means no notice period, no severance and no register of who lost what. Researchers produced the 40,000 figure because no employment record exists to produce it.

What to watch for next

Mark Graham, who studies digital labour at the Oxford Internet Institute, expects the same pattern elsewhere. Watch the shape rather than the warning. Companies sent this work abroad because labour was cheaper there, then automated it once the models could do it, and the second move took two years where the first took twenty.

On Monday morning British industry bodies were telling designers their jobs are safe. Bundi heard something similar in 2022. Nobody publishes reassurance for informal workers in Nairobi, and nobody counted them when they stopped earning either. We looked last week at how AI shortlisting misses the people companies actually hire. Nairobi is the other end of that machine.

This piece is informational, not a recommendation to buy, sell, or hold any asset.