If AI Does the Early Work, Where Does the Learning Come From?

AI can remove work that once occupied much of an early career. But some of that work also helped people develop the judgment and experience needed for what came next.

One of the promises of AI is that it can take work off our hands.

It can produce the first draft, summarize the research, analyze the information, prepare the report, write the code, organize the notes or identify patterns that once required hours of someone's attention.

Much of that sounds like progress.

But some of the work AI is particularly good at doing has traditionally been done by people early in their careers.

That raises a question we may need to spend more time thinking about:

If AI does the early work, where does the learning come from?

Some work produces more than an output

When we look at a task, we usually see the result it is supposed to produce.

The research produces an analysis. The analysis produces a recommendation. The draft eventually becomes the finished document. The review catches mistakes. The customer conversation produces information that shapes what happens next.

But work can produce something else at the same time.

It can change the person doing it.

The junior employee who works through the research begins recognizing what matters and what doesn't. The person preparing the first draft starts seeing patterns in how ideas fit together. Someone reviewing numbers repeatedly begins noticing when something doesn't look right.

The output matters, but so does what the person becomes capable of seeing and doing through producing it.

Early work has often been developmental work

Think about how expertise develops in many professions.

People usually don't begin with the most difficult decisions. They start with smaller pieces of the work.

They research. Prepare. Observe. Draft. Check. Revise. Make mistakes that someone more experienced catches. They see how their work is used downstream. Gradually, they encounter enough situations to recognize patterns and make better judgments themselves.

Not every early-career task is valuable. Some of it is tedious, unnecessarily repetitive or the result of processes that should have changed years ago.

We shouldn't preserve inefficient work simply because previous generations had to do it.

But before removing it, it may be useful to understand whether anything besides the output was being produced.

Efficiency and learning don't always ask the same question

AI encourages a reasonable question: Can this be done faster or better?

Learning introduces another one: What does someone learn by doing it?

Those questions won't always lead to the same answer.

An experienced professional may be able to use AI to accelerate work because they already have enough knowledge and judgment to evaluate what comes back. They recognize a questionable assumption. They know when something important is missing. They can distinguish an unusual answer from a wrong one.

Someone earlier in their development may be able to produce the same output with AI without yet having developed the judgment needed to evaluate it.

The result can look remarkably similar.

The capability behind it may not be.

Knowing how to use AI is itself a skill

There is clearly learning involved in using these tools well.

People need to learn how to frame a request, provide useful context, work iteratively, evaluate sources and understand where a tool is likely to be useful or unreliable.

Those are increasingly valuable skills.

But learning to use AI doesn't automatically replace the learning that occurred through the work AI now performs.

If someone can generate an analysis without having developed much understanding of the underlying information, the challenge isn't that they used AI. The challenge is knowing how they will develop the judgment required to recognize whether the analysis is any good.

That judgment has to come from somewhere.

We may need to design learning back into the work

For a long time, organizations haven't necessarily had to think explicitly about every part of this.

Some capability developed simply because doing the job required people to spend time doing the work.

If AI changes that pathway, development may need to become more intentional.

Perhaps an early-career employee uses AI to produce a first analysis but is expected to explain the assumptions behind it. Perhaps they compare their own judgment with the AI output before seeing the answer. Maybe they spend more time observing experienced colleagues make difficult decisions, following work downstream or reviewing cases where the obvious answer turned out to be wrong.

The answer won't be the same for every kind of work.

What matters is recognizing that removing a task and replacing the learning that occurred through that task are two different design decisions.

There may also be an opportunity here

AI doesn't only remove developmental experiences. It may create new ones.

If technology reduces the time people spend on routine preparation, perhaps they can encounter more situations, receive feedback sooner or spend more time understanding why a decision was made.

Someone who once spent hours assembling information might instead have more opportunity to discuss what the information means. A junior employee might be able to explore several approaches to a problem rather than having time to produce only one.

Used thoughtfully, AI could accelerate parts of learning rather than diminish them.

But that outcome isn't automatic.

It depends on whether we design the work only around producing the output more efficiently or also around developing the person who will eventually be responsible for more difficult work.

What are people learning by doing the work?

As AI becomes part of more jobs, we will understandably pay attention to which tasks it can perform.

There is another question worth keeping alongside that one.

What was someone learning by doing this work themselves?

If the answer is very little, perhaps we have found work that technology should remove.

If the answer includes judgment, pattern recognition, context, problem solving or understanding how the work fits together, then removing the task may require us to think about how those capabilities will develop another way.

Every job contains work that produces results and work that develops the person doing it.

Sometimes it's the same work.

As AI changes one, we should probably pay attention to what happens to the other.