Julie Yack

How do we nurture juniors into seniors when AI is doing the heavy lifting?

Not junior employees. Junior work.

AI may not eliminate junior employees. But it’s absolutely coming for junior work.

And I’m not sure we’ve thought enough about what else disappears when that work does.

The first drafts. The basic research. The simple code. The meeting summaries. The routine analysis. The work that takes time but doesn’t necessarily require years of experience to complete.

It’s exactly the kind of work AI is getting really good at.

And that’s a good thing.

I don’t think we should preserve tedious work just because the rest of us had to do it. Nobody gets extra credit for suffering through a task a machine can now accomplish in seconds.

But I’ve started wondering what else was happening while we were doing all that work. Because that was also how we learned.

We did something. Someone more experienced reviewed it. We got some of it wrong. We did it again. Eventually, we stopped making the obvious mistakes and started making more interesting ones.

We recognized patterns. We developed instincts. We learned that two answers could both be technically correct and still not be equally good. Eventually, we developed judgment.

And eventually, we became the people reviewing someone else’s work.

That’s how juniors became seniors.

So what happens when AI does the work we used to learn from?

This isn’t the first time technology has taken over something humans used to do.

Calculators can do arithmetic better than most of us. We still teach math. GPS can tell us exactly where to turn. We still think there’s value in understanding where we are and where we’re going. Spellcheck has been correcting our spelling for decades. We still teach people how to write.

We’ve always had to decide which fundamentals humans still need after technology makes them unnecessary for getting the work done.

AI dramatically raises the stakes.

Because it’s not just calculating the answer or correcting the spelling.

It’s doing the research. Writing the first draft. Analyzing the information. Writing the code. Making recommendations.

It’s doing many of the things we’ve traditionally handed to someone early in their career so they could learn how the work gets done.

I’ve said for a long time that you should have some idea what AI is going to tell you before you ask. Not the exact answer. But enough context to recognize whether the answer makes sense.

I’ve been doing the kind of work I do for decades. I have that context.

How does someone new get it now?

We often say that humans need to review AI’s work. Keep the human in the loop. Check the output. Validate the answer.

But how?

How do you know whether the answer is good if you’ve never done the work yourself? How do you know what’s missing? How do you recognize the edge case?

How do you know when it’s a house of cards if you’ve never played poker?

And the answers that worry me aren’t usually 2+2=5.

Most knowledge work isn’t that binary.

There may be three perfectly reasonable approaches to a problem. All three might work. Experience is what tells you that one of them is a terrible idea here. In this organization. With these people. Under these constraints.

That’s not fact checking.

That’s judgment.

Maybe part of the answer is teaching people to use AI to challenge AI. Ask it to expose its assumptions. Consider alternatives. Find weaknesses in its own recommendation. Tell us what might change the answer.

But even then, who decides whether the challenge was good enough?

The easy answer is that humans should spend more time reviewing what AI produces. I don’t think that works either.

If AI saves me 80 percent of the time doing the work and I then spend that 80 percent validating everything it did, we’ve missed the point.

Speed matters.

Maybe we need to teach people when to question AI rather than teaching them to question everything. Maybe some foundational work still needs to be done without AI. Maybe AI itself becomes part of the apprenticeship. Maybe we need to spend less time teaching people how to produce the work and more time exposing them to the decisions surrounding it.

I don’t know.

For a long time, juniors got junior work because they were juniors.

Some of that work taught them something important. Some of it gave them repetitions they needed to recognize patterns later. Some of it exposed them to situations that eventually became experience. And some of it was just work the senior people didn’t want to do.

We bundled all of that together and called it experience.

People eventually became senior.

Now AI gives us a reason to pull that bundle apart.

Which parts actually mattered?

Which repetitions created judgment?

Which mistakes were worth making?

Which experiences taught someone to recognize the difference between an answer that works and an answer that’s right for the situation? And which parts were just grunt work we can happily never ask another human to do again?

We are moving incredibly quickly to figure out which work AI can do for us.

We should.

I don’t want people spending hours doing something AI can accomplish in seconds simply because that’s how I learned. But I also don’t want us to assume expertise will magically appear because the work got easier.

Today’s senior professionals had years to develop judgment while technology was less capable. The people coming next are going to develop theirs in a completely different environment.

Maybe the question isn’t how much junior work we’re comfortable giving to AI. Maybe it’s which parts of that work were quietly turning juniors into seniors. And whether we know how to replace them.

So I keep coming back to the same question.

How do we nurture today’s juniors into tomorrow’s seniors when AI is doing the heavy lifting?

I don’t have the answer. I’m not sure any of us do yet.

But if we’re going to fundamentally change how people work, we’d better also start thinking about how they learn.