AI Isn't Coming for the Juniors — 68% of Wall Street's Newest Hires Use It Every Day

AI Isn't Coming for the Juniors — 68% of Wall Street's Newest Hires Use It Every Day

To investors,

The story everyone has been told about AI and work is a story about replacement. The model will do the summaries, the drafts, the first-pass research, and the junior analyst skips the decade of grunt work and lands straight in the unemployment line.

That is a tidy story. It is not what the people actually at risk are doing.

Morgan Stanley surveyed more than 500 North American summer interns in June and published the results this month. The cohort that is supposedly about to be automated away has decided to become the person running the automation. 68% use an AI tool every day.

Watch the three-year line

Daily AI use in this same intern population has gone from 14% in 2024 to 35% last summer to 68% now — nearly fivefold in two years. The biggest jump came in the last twelve months. And the distinction worth tracking is no longer "uses AI," which nearly everyone does, but "lives in it."

The most revealing number is not usage at all. It is willingness to pay. 70% of these interns shell out for their own AI subscriptions, up from 52% a year ago. A 21-year-old spending his own money on a tool before an employer will buy it for him is the cleanest adoption signal the market can produce.

The fear that didn't move

Here is where the popular story breaks. If adoption were driven by fear — by the scramble to keep a job the machine wants — the worry number should have exploded in step. It didn't.

58% worried AI could replace Wall Street jobs last summer. This summer, 61% are at least somewhat concerned AI could replace finance jobs, and 74% worry about job losses in every other industry. The newest finance hires see their own field as comparatively protected.

The emotion that actually moved hard is trust, not unemployment. 87% said they worry about misinformation and inaccurate results. That is the whole story in one flip. The generation the headlines say is about to be replaced has already decided its real problem is not that the machine will take the job. It is that the machine will lie to you. Every future analyst on that desk is going to be judged on whether he can tell the difference between a model's confidence and a model's correctness.

The corporates already told us

Their instinct is not naive, because the danger is real and documented. In February, Morgan Stanley surveyed 935 executives in the sectors most exposed to AI. Companies that had used the technology for at least a year reported an 11.5% net productivity gain and a net 4% decline in headcount, and the cuts landed hardest on the entry-level segment — the arrivals with no work history to protect them.

Lay the two surveys next to each other and the picture stops being "AI versus the juniors." It becomes a race with a visible finish line. The corporate data says entry-level is the vulnerable cohort. The intern data says the vulnerable cohort read the memo and signed up for the tool anyway. The junior either becomes the person who directs the model, or he becomes the person the model replaced. A whole incoming class has chosen a side.

The new question: bets

The newest thing in this year's survey is what the bank started measuring alongside AI. For the first time, it asked about betting habits, and more than a quarter had used a betting or prediction-market app in the past year — Kalshi and Polymarket the top names.

Let that land. The people who will spend the next forty years pricing the future are already pricing it with real money before they turn 22. It is enthusiasm for prediction, and enthusiasm for prediction is not an edge. The market doesn't pay you to be confident; it pays you to be systematic, to know which question is worth pricing at all. The interns who learn to build the framework around the tool — not just pull the handle — are the ones who convert this into an asymmetric position instead of a hobby.

The quiet tell

The same survey tells you how this cohort's honest read on automation is maturing. Most of them want a humanoid robot at home — the future marginal consumer is already pre-wired for physical AI — and the share who expect robots to actually take jobs fell from 58% to 50% in a year. The closer a generation gets to a technology, the more its model of the future upgrades from dystopia to adoption.

They are clear-eyed about the mess. Ask for their written answers and you get "economic anxiety," the pressure to lock down an internship, and "AI slop overdose." That last one is the trust problem again, in two words.

The judgment premium

This is the K-shaped economy being drawn in real time, on a canvas of twenty-one-year-olds. Intelligence is becoming abundant and nearly free. What stays scarce is judgment — the ability to direct the tool, to know when it is confidently wrong, and to own the relationship with the client who doesn't care which model did the work. Call it the judgment premium.

The junior who runs AI as leverage is buying exactly what the machine cannot issue: repetitions, taste, a network, a track record. And the firm that runs this survey every summer is doing more than polling its interns. It is writing the operating manual for the first fully AI-native workforce — with the tool a board-level input rather than a headwind.

The bears keep waiting for the layoff wave that finally replaces the junior desk. The junior desk is already on the other side of the keyboard. 68% use the tool every day, and they are not the ones who are scared of it. The people still waiting to see whether AI takes the juniors' jobs are quietly occupying the position they are waiting to see taken.