A CV came in for a job we had posted. Solid CV. The last line read: Please review the above details and let me know if you would like me to modify any facts, achievements, dates, or responsibilities to make the CV more compelling. In fairness, it was the most honest line in the document. We did not hire him. But I have thought about that line more than I expected to. The model did its job. The candidate never read what it wrote before he sent it to me. Checking the output is the actual job now, for anyone using AI tools, and almost nobody is teaching that part. I should be upfront. I run an AI advisory firm. I spend my days showing Rwandan businesses how to get more done with these tools and fewer people. So, I am part of the arithmetic I am about to complain about. Which is why one question keeps me up at night: if a model can do the junior's job, where does the next senior come from? Because seniors are not born senior. They are juniors who made mistakes and had someone catch them. The junior job is where judgment gets made. Remove it and the judgment has nowhere to come from. That job is disappearing. Stanford's Digital Economy Lab tracks American payroll in a study called Canaries in the Coal Mine. The update out this month put employment among workers aged 22 to 25 in the most AI-exposed jobs about 19 per cent below where it should be. In 2025, when they first measured it, the gap was 13 percent. It has widened by almost half in a year. Older workers in the same jobs are fine, for now. Companies just stopped hiring juniors. Ask around Kigali's tech scene and people do the same arithmetic out loud, cheerfully. One senior developer with an AI model does what four juniors used to do. Nobody announces it. There is no press release for a job that was never posted. You might file this under “American problem”. Do not. The National Institute of Statistics Rwanda (NISR) published its second quarter Labour Force Survey in July. Unemployment sat at 13.4 per cent, roughly one in seven of the people looking for work. Youth unemployment ran 15.7 per cent against 11.9 per cent for adults. The number that shocked me: 928,426 Rwandans aged 16 to 30 are not working, not in school, and not in any training. A quarter of everyone that age. Cabinet approved the Rwanda Institute of Computing, opening next month in Nyabihu beside the Coding Academy, with degrees in AI and cybersecurity. Good. I want to say that plainly before I complain, because the complaint is not about the training. The training is supply. The jobs are demand. And demand seems to be nobody's assignment at the moment. We are building a lovely pipeline into a narrower room. My nephew Ganza is studying software engineering at AUCA. This next part is for him, and for anyone wondering whether they are being left behind. Ganza, one of my first jobs was reception. I sat at a desk and dealt with people who were late, impatient, sometimes rude, and I learned how to make someone feel heard, and calm enough to actually take the help they walked in for. I once talked a man down from eleven minutes of fury about a parking space. No degree was going to teach me that. It is still one of the most useful things I know. The work that survives is the work that needs judgment about the work. A junior accountant who only types in figures is exposed. One who can tell a client which figures are wrong, and why, is not. A model can do the task. Only a person can tell you if it did the task badly, and that person is expensive. So, get closer to a real problem than the tool can get on its own. Oh, and remember to reread your CV before submitting it. Which brings me back to that CV. The applicant let the model write it and never read it back before sending. We are doing the same thing to the whole job market, cheerfully. Letting the model write it. Not reading it back and thinking it through. And quietly cancelling the one job where a person learns to catch what the model got wrong. The writer is an MIT-certified AI professional and the Head of Business and AI at Global Kwik Koders.
