The New York Times published a report on September 5 about Nairobi, Kenya, and The Decoder retold it two days later. The story covers a business that has existed locally for more than a decade: writing coursework essays for international students at American and British universities. At the industry's peak, at least 40,000 people worked in it.
The writer featured in the report is Teresios Bundi. Over twelve years he wrote more than 2,500 papers, charging $40 to $70 each, taking on assignments in medicine, computer science, and engineering. The business was never a fringe pursuit locally: roughly 80% of employment in Kenya falls under the informal sector, and the government has promoted online gig work since 2016, putting essay writing right at the intersection of English fluency and a workable time-zone gap.
Orders started dropping in 2022
Once ChatGPT launched, international students no longer needed to outsource their assignments to a stranger in East Africa. Order volume and prices fell together. Some writers shifted to a new line of work: rewriting AI-generated text so it slips past school detection tools, known in the trade as "humanizing." This new business has a much lower technical bar to entry, and it pays much less.
It's not just essay writing that's shrinking
The report groups a few other jobs alongside essay writing: transcription, data labeling, and content moderation, which together have been the mainstays of Kenya's online gig economy in recent years, and whose orders are shrinking too. They ended up in East Africa for the same reason essay writing did — cheap English-speaking labor and a workable time difference — and now they're contracting for the same reason. Oxford's Mark Graham predicts in the piece that similar disruption will repeat elsewhere.
"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."
Coming from someone who just lost his livelihood, that line carries different weight than it would from a conference stage. His industry wasn't shut down by policy, and it wasn't undercut by a cheaper competitor — the demand side simply vanished: international students suddenly had a free, instant substitute.
A case with a clean boundary
Most research into AI's effect on employment comes down to statistics: hiring for a given role down by some percentage, the wage premium for some skill shrinking by some amount. What makes the Nairobi case notable is how clean its boundaries are — one business, one city, one industry whose headcount can actually be counted, with a start and an end point that can both be marked. The 40,000 figure is an industry estimate with no official registry to verify it; even discounted, the order of magnitude still holds up.
What the report doesn't answer is where these people went. The "humanizing" path is itself unstable, since it depends on detection tools lagging behind generative models; Kenya's 2016 online-gig push hasn't offered a new outlet either. For now, only the two endpoints can be confirmed: a workforce that once numbered 40,000, and whatever scattered orders remain today.
Sources: The New York Times, CocoLoop, The Decoder; figures on the 40,000-person workforce, $40-70 per-paper rates, the 2,500-paper output, and Kenya's 80% informal-employment share are checked against the reporting.