The Last Rung of the Ladder

We usually talk about artificial intelligence and work as though the central question were simple:

Which jobs will disappear?

That matters, but there is another question hiding underneath it:

How will people learn to become good at the jobs that remain?

For generations, expertise has had a ladder.

You began with work that was slower, narrower, and sometimes painfully repetitive.

You checked the figures.

You summarized the meeting.

You reviewed the documents.

You fixed the small bugs.

You prepared the first draft.

You watched someone more experienced tear it apart.

Then you did it again. None of this looked glamorous; much of it looked inefficient.

It was also how judgment was built.

The Work Beneath the Work

Artificial intelligence is especially good at many of the tasks we have traditionally given to beginners.

Draft this.

Summarize that.

Compare these documents.

Research the background.

Clean the spreadsheet.

Write the first version.

Find the obvious error.

Those are precisely the tasks organizations are eager to automate.

The economic logic is easy to understand.

If a machine can complete in thirty seconds what once took a junior employee three hours, why pay for the three hours?

But those three hours were not only producing a document.

They were producing a worker.

That distinction is easy to miss.

Entry-Level Work Was Never Only About Output

A young lawyer reviewing hundreds of contracts is not simply moving text from one place to another.

Over time, patterns become visible.

Certain clauses begin to feel wrong before the lawyer can fully explain why.

A junior programmer fixing small defects begins to understand where systems tend to break.

A young journalist checking names, dates, transcripts, and source material slowly develops a sense for what does not add up.

An analyst building ordinary models begins to recognize when an extraordinary result is probably a mistake.

This is not magic.

It is exposure.

Judgment grows from repeated encounters with reality.

The beginner sees individual cases.

The experienced person eventually sees structure.

AI can accelerate the casework.

But if the casework disappears entirely, we should ask what happens to the structure that used to grow inside the person doing it.

The New Entry-Level Paradox

Employers are already beginning to ask more of people entering the workforce.

Recent labor-market research suggests that entry-level roles exposed to AI increasingly demand skills once associated with more experienced workers: judgment, leadership, creativity, and the ability to interpret rather than merely produce information.

That sounds promising.

Who would object to better judgment?

The problem is obvious once you state it plainly:

Where is a beginner supposed to get senior judgment?

We may be creating jobs that require experience while automating the work through which experience was traditionally acquired.

This is the new entry-level paradox.

The first rung of the ladder is disappearing at the same moment we are asking young workers to climb higher.

Apprenticeship Was Hidden Inside the Job

Modern offices rarely use the word apprenticeship.

But much of entry-level work has always functioned that way.

You sit near people who know more than you.

You perform tasks of limited consequence.

You make mistakes while the consequences are still containable.

Someone corrects you.

You begin to understand not just the rule, but when the rule fails.

Eventually, responsibility increases.

That is apprenticeship whether we call it that or not.

The danger is not that AI will eliminate drudgery.

Some drudgery deserves to disappear.

The danger is that we eliminate the learning architecture along with the drudgery.

We automate the assignment and forget what the assignment was teaching.

Efficiency Can Consume Its Own Future

Imagine an organization that becomes exceptionally efficient.

Its senior people use AI to perform work once delegated to juniors.

The juniors who remain use AI to skip much of the mechanical learning that earlier generations endured.

Productivity rises.

Costs fall.

Everyone celebrates.

Ten years later, the organization needs new senior people.

Where do they come from?

The old experts retire.

The junior workers have spent years supervising systems they did not build through the same accumulation of mistakes, corrections, and repetition.

They may be extremely capable.

They may also have gaps nobody noticed because the machines were good enough to cover them.

An organization can become so efficient at removing beginner work that it eventually removes its own method of producing experts.

That is not efficiency.

That is borrowing competence from the past.

This Is Not an Argument for Pointless Work

We should be careful here.

There is nothing noble about making young people perform useless tasks simply because older generations had to perform them.

Bad apprenticeship is still bad work.

There are plenty of repetitive jobs that teach almost nothing.

AI should remove some of them.

The goal should not be to preserve inefficiency.

The goal should be to preserve formation.

If a machine can write the first draft, perhaps the beginner should spend more time explaining why the draft works.

If AI can find the obvious software bug, perhaps the junior programmer should trace how the bug entered the system.

If a model can summarize the contract, perhaps the young lawyer should compare its summary against the source and defend which clauses matter.

The task can change.

The learning cannot disappear.

We May Need Deliberate Apprenticeship Again

For a long time, companies could treat training as a side effect of employment.

Hire enough junior people.

Give them work.

Some become experienced.

AI may make that model unreliable.

If so, organizations will have to become much more deliberate about creating expertise.

That could mean longer residencies.

Structured rotations.

Supervised decision-making.

Protected practice.

Human review sessions in which the point is not merely to correct the AI output but to make the junior employee explain the correction.

It may even mean giving beginners some tasks that an AI could perform faster because the human still needs to learn what the task teaches.

That sounds inefficient.

Education often is.

A flight instructor does not land the aircraft simply because the instructor can do it better.

The student eventually has to touch the controls.

The Real Skill Is Not Prompting

Much of today’s advice to young workers is some version of:

Learn to use AI.

They should.

Ignoring the tools would be foolish.

But knowing how to operate an intelligent system is not the same as developing the judgment required to evaluate it.

The worker of the future will need both.

They will need to know how to ask.

And how to doubt.

How to accelerate.

And when to slow down.

How to delegate.

And what must never be delegated without review.

Those abilities are unlikely to appear automatically because someone learned how to write a better prompt.

They have to be cultivated.

Someone Still Has to Become the Expert

The future of work is not simply a contest between humans and machines.

It is also a question of generations.

Today’s experienced workers learned in a world where the machine could not do much of the beginner’s work.

Tomorrow’s workers will not have that luxury.

That means institutions have a responsibility they have barely begun to recognize.

If we remove the lower rungs of the ladder, we cannot simply tell the next generation to jump.

We have to build another way up.

Because eventually the person supervising the machine will need to know when the machine is wrong.

And somebody has to teach them how.

Dale Joseph is the author of Thought Partners: Preserving Cognitive Sovereignty in the Age of AI and founder of the Emergence Institute. He worked for years as a consultant helping install hospital networks before turning to writing and systems thinking. He lives in Boynton Beach, Florida.