Who Learns Finance When AI Does the Work?

Work in progress

A lot of the foundational work in finance isn't particularly glamorous. Building models, reconciling messy data, tracing numbers back to their sources, and figuring out why something doesn't add up.

But that work has traditionally served two purposes. Companies needed the analysis, and junior professionals learned by producing it.

AI is beginning to separate those two things.

The early employment data is worth paying attention to. An October 2026 analysis from Revelio Labs found that junior headcount at U.S. banks had fallen 4.7% since October 2022, while senior headcount continued growing. Stanford researchers found a 19% relative employment gap for workers aged 22 to 25 in highly AI-exposed occupations compared with less-exposed occupations.

Neither finding establishes that AI caused these changes, and neither proves that finance professionals are becoming less capable. Banking isn't the entirety of finance, either.

Still, the direction raises a question worth taking seriously.

If the work that once trained junior professionals becomes increasingly automated, how will the next generation develop the judgment needed to challenge it?

The training dilemma

Imagine you're running a finance organization. Suppose an experienced analyst equipped with AI agents can now handle work that previously required several junior analysts.

You could continue hiring and developing junior talent. Or you could run a leaner team, automate more of the work, and recruit experienced people when you need them.

The second option might make perfect economic sense for your company.

But what happens when everyone makes the same decision?

Consider two companies facing that choice.

This resembles a prisoner's dilemma. Each company might prefer to avoid training costs regardless of what its competitor does, even though both could be better off if they invested in developing talent.

No one has to make an irrational decision for the collective outcome to be worse.

Of course, the analogy depends on some assumptions. Companies must benefit from expertise developed elsewhere, and the industry must still need that expertise in the future. If AI reduces demand for experienced finance professionals enough, a smaller talent pipeline might be entirely rational.

That's what makes this problem interesting.

A different kind of apprenticeship?

I'm not convinced the answer is to preserve the old way of learning finance.

There was value in building models by hand. You learned how financial statements connected, where assumptions entered the analysis, and how easily a small mistake could distort an apparently reasonable conclusion.

But not every hour spent reconciling a spreadsheet was an hour of valuable education.

What if junior professionals could spend less time producing the analysis and more time challenging it? They might learn through questioning forecasts, investigating the economics of a business, participating in actual decisions, and revisiting those decisions once the outcomes become clear.

AI might even accelerate that process.

The risk is that someone becomes very good at prompting an agent to produce a model and a persuasive explanation without developing the understanding needed to recognize when either is wrong.

And there's a second problem. Even if we design a better apprenticeship, who pays for it?

If companies can get the deliverable without training the person, developing future expertise becomes a deliberate investment rather than a byproduct of completing necessary work.

I'm still working through where this leads. Perhaps AI makes financial judgment more accessible and reduces the need for dedicated finance professionals. Perhaps it raises the value of the experienced people who remain. Both could happen at once.

But if we still need people capable of questioning the models and narratives AI produces, we'll need to think much more deliberately about how those people learn.