Last week, I spoke to a room of students at the University of the Western Cape. I was not there in person. I was in Kigali, joining virtually from a city that has spent thirty years rebuilding itself from the ground up, which felt, in retrospect, like exactly the right place to be talking to young Africans about the future of work.
The session was billed as “Using AI for your job search.” I took it somewhere slightly different.
I told them there are two kinds of graduates leaving universities right now. The ones the machine works on. And the ones who work the machine. My aim for those thirty minutes was to move every person in the room from the first group to the second. Not by inspiring them, though I hope I did that too, but by giving them four specific, practical moves they could run this weekend.
I want to share those moves here, and then I want to tell you what I think they are actually about, because the deeper argument is the one that matters most.
The Four Moves
The first move is CV translation.
Your CV has a problem. It was written for a human reader. But in this era, the first reader is almost never human. It is an Applicant Tracking System, a piece of software that scans your CV for specific phrases from the job description and decides, in seconds, whether you move forward or not. The second reader, if you make it past the first, is a human recruiter who will spend an average of seven seconds on your CV before deciding whether to read further.
Most graduates do not know this. They spend hours polishing the prose of a document that will be read by a machine before it is read by a person. So the first move is not to rewrite your CV. It is to translate it.
Here is the prompt. Paste the job description and your current CV into any AI tool of your choice and say: identify the exact phrases from this job description that are missing from my CV but that I have real experience to back up. For each one, suggest where in my CV to add it without exaggerating my experience.
What this does is surface real claims you already have, but have not stated in the language the system is scanning for. The CV stays yours. It just speaks the right dialect.
The second move is application targeting.
The biggest mistake I see from job seekers, and I say this as someone who runs a talent company, is focusing on volume. Fifty generic applications will lose to five thoughtful ones, every single time. AI does not exist to help you send more applications. It exists to help you send better ones.
Before applying to any company, ask AI to summarise that company’s recent activity, news, product launches, leadership changes, recent annual reports. Then ask: based on this summary, what is the most likely problem the team I am applying to is trying to solve right now, and based on my experiences, where can I credibly position myself as someone who can help solve it.
Here is the mindset shift underneath this move. Employers are not hiring you because they want to give you a job. They are hiring you because there is a vacancy in their value creation process, a specific problem they need solved, a specific gap in their capability. Your application is not a presentation of yourself. It is a response to their problem, with you as the solution. The moment you understand this, your cover letter stops being about you and starts being about them, with you as the answer.
The third move is interview rehearsal.
Most candidates rehearse the questions they hope they will get. The move that separates the ones who win from the ones who almost win is rehearsing the questions you are afraid of.
Paste the job description and your CV into AI and say: roleplay as a skeptical interviewer for this role. Ask me the five questions most likely to expose the weakest claims on my CV. After I answer each one, give me honest feedback on whether my answer was credible.
One important instruction I give every student I work with. Do not type your answers. Speak them out loud and use the microphone. Interviews are spoken. If you only ever practice in writing, your mouth is unprepared. The discomfort of doing this exercise in your room is far cheaper than the discomfort of feeling it for the first time in front of a hiring panel.
The fourth move is the audit.
This is the move almost nobody teaches, and it is the most important one.
AI hallucinates. It fabricates. It produces beautiful, confident, wrong answers. And in a job search, the cost of an unchecked claim does not show up at the application stage. It shows up in the interview room, when you cannot defend something the machine wrote on your behalf.
Audit AI output on two levels. First, factual correctness. Triangulate across tools. Ask the same question on Gemini and on ChatGPT and see where they diverge. Where they agree confidently and are still wrong is the category that will catch you, so verify anything that will be tested.
But there is a second level of auditing that is harder and more valuable. AI does not know your story. It does not know that you built a sales tracking system for your mother’s market stall when you were seven, and that this is where your instinct for data analysis came from, long before you ever studied statistics at university. It does not know the specific texture of your life, the problems you have already solved, the cultural context you carry, the way you understand a room that someone who did not grow up where you grew up simply cannot. These things are invisible to the machine. They are yours. And they are the most powerful things you can inject into any application, any interview, any piece of work.
Audit means: take what the machine gives you, verify it, and then make it human.
The Argument Underneath the Moves
Those are the four moves. Translation, targeting, rehearsal, audit. You can run all four this weekend.
But I want to tell you what I think they are actually about, because the practical case and the philosophical case are the same argument.
The phone in your pocket right now has 100,000 times more transistors than the Cray-1, the most powerful computer in the world in 1976. On raw computing power, it is roughly 20,000 times more capable.
So here is the question I put to the UWC students yesterday, and I put it to you now. If you use your phone to write a memo, run an analysis, send fifty emails, are you 20,000 times more productive than someone doing the same work in 1976?
Obviously not.
So why does your productivity lag so far behind the power of the tool in your hand.
Because tools alone do not produce productivity. The Cray-1 needed humans to decide what to compute, why it mattered, and what to do with the result. Your phone needs humans to decide what to write. AI needs humans to decide what is worth building. The bottleneck has never been the silicon. It has always been the human capacity to direct it.
In any system, the throughput is set by whatever does not scale. Compute is scaling. What is not scaling, what cannot be automated or compressed or cheapened, is the capacity to bring genuine judgment, presence, and contextual intelligence to a situation. The capacity to read a room. To understand a client in a way a model trained on data from another country, another culture, another set of lived experiences, never will. To know when a problem needs a workflow and when it needs a human voice.
This is the description of what the market is now paying for most. The Microsoft and LinkedIn 2024 Work Trend Index found that 71% of leaders globally would rather hire a less experienced candidate with AI skills than a more experienced one without them. The market is not looking for the person who can use the tool. It is looking for the person who can direct the tool and then do the thing the tool cannot do.
That is what the four moves are training you toward. Not to just use AI faster. To use AI so that the part of you the machine cannot touch, your judgment, your story, your cultural intelligence, your presence, can finally be the part that the market pays for.
To the Young African Specifically
The story you are being told about AI and the future of work is half true and entirely misleading. Yes, AI is absorbing whole categories of work, especially those that can be specified. That is happening and it will continue. But the other half of the story, the half almost nobody is telling you, is this.
As AI absorbs what can be specified, what cannot be specified rises in value. Judgment rises. Taste rises. Context rises. The capacity to feel what a situation actually needs. To understand a Cape Town client, a Lagos founder, a Nairobi team, in a way a model trained on California data never will.
This is Africa’s advantage. Raw, abundant, largely unrecognised by the continent itself, but structurally real. The youngest population on earth. A generation that learned to do more with less before that became a business model. That coordinated across borders because they had no choice. That built on constraint until constraint became fluency.
The question is not whether the machine is coming for the work. It is whether you will be the kind of worker the new economy cannot replace.
You are not learning AI in order to keep up. You are learning AI so that the part of you the machine cannot touch can finally be trained, nurtured, and brought fully into the economy.
AI is not here to replace your work. It is here to restore your worth.
