Ask a senior professional how they became good and you will often hear about work they would never want to do again.
The first drafts.
The basic research.
The spreadsheet cleaning.
The repetitive coding.
The meeting notes.
The small client requests.
The hours spent discovering why something that looked simple was not.
Much of that work was inefficient. Some of it was genuinely tedious. AI can now remove a growing portion of it.
That sounds like progress.
It also creates a training problem we have barely begun to solve.
Entry-level work has traditionally served two purposes at once. It produced useful output for an employer, and it exposed a beginner to enough repetition that patterns slowly became judgement.
A junior accountant reconciles transactions and begins to notice what looks wrong.
A junior developer fixes small bugs and learns how systems break.
A young lawyer reads documents and starts recognising what matters.
A new marketer writes weak copy, receives feedback and gradually understands audience.
The task is not the same thing as the learning. But the learning often hides inside the task.
Now imagine that AI produces the first draft, summarises the documents, writes the basic code, analyses the spreadsheet and prepares the meeting notes.
The senior employee becomes faster.
What happens to the beginner?
This is not an argument for preserving pointless busywork. Nobody should spend six hours manually doing something a reliable system can complete in six minutes simply because previous generations suffered through it.
The challenge is to separate unnecessary labour from necessary learning.
Companies that fail to make that distinction may create a strange organisation: highly productive senior people supported by AI, but fewer opportunities for juniors to accumulate the experience required to replace those seniors later.
That is not a hypothetical concern confined to Silicon Valley.
It matters especially in African labour markets where young populations are entering economies that already struggle to create enough high-quality entry points.
The worst response would be telling young professionals simply to “learn AI.”
AI literacy is useful, but it does not solve the apprenticeship problem.
A person can become excellent at prompting a coding assistant without understanding architecture. They can generate a polished financial model without recognising a bad assumption. They can produce persuasive copy without learning strategy.
Tools can hide ignorance remarkably well.
Until something goes wrong.
The future of junior work therefore needs intentional design.
One approach is to keep the responsibility while changing the mechanics.
Instead of asking a junior analyst to spend four hours producing a first pass manually, allow AI to produce it — then require the junior to verify every assumption, identify weaknesses, compare alternatives and explain the recommendation.
The task becomes less about production and more about judgement.
For developers, AI-generated code should create more opportunities for review, testing, debugging and architecture discussion, not less.
For writers, the learning may move from producing endless first drafts to editing, interviewing, fact-checking, voice and argument.
For designers, generation can increase the number of options, but someone still needs to understand hierarchy, context, usability and taste.
This changes what mentorship must look like.
Managers cannot assume that experience will accumulate automatically through exposure to repetitive tasks. They may need to create deliberate learning loops: shadowing, critique sessions, post-mortems, rotations, simulations and progressively harder ownership.
That requires time.
Ironically, the time saved by AI may need to be reinvested in teaching humans.
There is also a responsibility on juniors.
The temptation will be to use AI to look competent before becoming competent.
That is dangerous because careers are long and reality eventually asks questions the interface cannot answer for you.
A useful personal rule is simple: never outsource the part of a task you are currently trying to learn.
If you are learning research, do some research before asking for synthesis.
If you are learning code, make sure you can explain what the generated code does.
If you are learning strategy, do not confuse a well-formatted answer with an insight.
Use AI after thinking, during thinking and to challenge thinking — not only instead of thinking.
Employers also need to reconsider what “entry level” means.
For years, job descriptions have demanded experience from people applying for roles supposedly designed to provide experience. AI could make that contradiction worse if companies use automation as an excuse to eliminate junior positions entirely.
A healthier model would treat early-career roles explicitly as capability-building positions with measurable progression.
The organisations that do this well may gain an advantage.
Because the future will still need senior people.
And senior people do not emerge from software.
They emerge from years of decisions, mistakes, feedback, context and responsibility.
AI can compress some of that journey.
It cannot eliminate the journey.
The companies celebrating the disappearance of junior work should therefore ask one uncomfortable question:
Who are we training to become excellent five years from now?