AI digital employees enter the headcount, and the historical mission of interns may be over

By | Xunkong's Marketing Revelations

I have a friend who works in finance. Their company used to hire one or two interns every year. This year, they bought two ChatGPT Pro memberships, didn't hire a single intern, and the full-time employees actually found their work easier than before.

When I heard about this change, my first thought was that the college students who might have gotten those internships probably won't ever know about it. They'll just feel that finding an internship got a little harder this year. This kind of job change doesn't even require a layoff—the position simply gets eliminated before the next round of hiring.

On October 29, I saw another piece of news about JD.com. Phoenix Technology, citing Big Tech Daily, reported that JD.com's finance CFO system was running a partial pilot in which AI digital employees were placed at the P3 job grade and given independent headcount.

The report mentioned that some Agents' work performance and direct supervisor evaluations had already surpassed those of interns.

One is a small team that stopped hiring, the other is a big company starting to assign job grades to digital employees. They both raised the same question for me: how much longer can the intern position survive?

To answer that question, we first need to look at why interns exist.

01 From Apprentice to Intern

What we call internships today includes some courses arranged by schools and some job opportunities students find themselves—they don't come from the same place. But young people learning a profession by doing the work has been going on for a long time.

In a medieval workshop, a boy of twelve or thirteen was sent to a master, signed a contract, moved into the master's house, and served a full seven years. He received no wages; the master provided food and lodging and taught him the craft. In the end he had to produce a qualified finished piece before he could become a journeyman, and only after that was he qualified to open his own workshop.

The terms were simple: the master got a pair of cheap hands, and the apprentice got a craft he could live on, plus the qualification to enter the trade.

But guild apprenticeship was never an ideal school. Guilds limited the number of apprentices and also limited how many could open businesses. Craft was a scarce resource, and scarcity itself was the source of bargaining power. Much of the work apprentices did over those seven years had nothing to do with training—it was just the chores the workshop needed.

Later, schools took on more and more educational functions, but people found that some skills were still hard to learn from classes alone. Medicine especially: there is a long distance between reading a textbook and facing a real patient.

From the late 19th to the early 20th century, American hospitals gradually developed a more systematic residency training system. Medical school graduates entered hospitals and, under the guidance of senior doctors, took part in ward rounds, wrote medical records, and performed clinical procedures. This was an important transition from classroom learning to independent practice.

In 1906, Herman Schneider of the University of Cincinnati sent engineering students into factories, alternating classes and work—cooperative education began here.

Although these systems differ, they all solve the same problem. Between the knowledge the classroom can teach and the ability to truly work independently, there is still a stretch of road that someone needs to guide you along.

Today's offices have compressed that road even shorter.

In industries like finance, advertising, and media, newcomers usually start with gathering materials, checking data, organizing meeting notes, and preparing first drafts. These tasks aren't highly technical, but they are the newcomer's first channel to understanding the business, and companies also use them to observe a person's learning speed, reliability, and way of collaborating.

There is a very typical training scene in The Devil Wears Prada. Andy has just joined the magazine and smiles at two belts that look almost identical; Miranda immediately explains the difference between them to her.

What this scene really shows is how a workplace newcomer gradually comes to understand professional standards they originally couldn't see. Later, Andy begins to anticipate tasks and understand the rhythm of the work—not through an operations manual, but by observing, making mistakes, receiving feedback, and gradually growing through one concrete task after another.

So the core of the internship system is the dual identity of the same basic work. The same task is a deliverable for the company and a learning opportunity for the newcomer.

Someone has to explain the business, break down the tasks, and check the results. What newcomers submit often needs repeated revision. By the time they finally get familiar with the work, the internship may be almost over.

As a result, companies' demand for inexperienced junior employees will decline.

An August 2026 study by the Stanford Digital Economy Lab found that AI is affecting young people's job opportunities. In industries easily affected by AI, the employment level of young people aged 22 to 25 was about 19% lower than what it would have been under previous trends.

More notably, this effect mainly shows up as companies reducing hiring of newcomers, rather than large-scale layoffs of older employees.

For job seekers, the trouble is that the experience companies want has to be gained somewhere. In the past, you could go in first and do some basic work, learning on the job if your skills weren't enough. Now, companies may want you to be able to handle a task independently from your first day.

But does that mean interns will disappear entirely? Not necessarily—JD.com still announced in March this year that it was opening more than 10,000 internship opportunities that could lead to full-time positions. At least, a big company deploying digital employees and keeping its internship hiring plans can happen at the same time.

Companies still need to select people and build a talent pipeline, and some work still requires real people.

03 Knowing How to Use AI Is Still a Long Way from Knowing How to Work

In the National Association of Colleges and Employers (NACE) spring 2026 survey, of 159 employers who answered the relevant questions, 57.8% said they would assign interns projects that require the use of AI tools and skills. For these companies, knowing how to use AI has already become part of the internship.

For example, doing an industry report. AI can quickly provide the market size, competitive landscape, and development trends, and it even has the final recommendations ready. An experienced person looking at the report might first check whether two data points use the same statistical basis, and whether the so-called market growth comes from increased sales volume or from higher prices.

A newcomer may not think of these things. What he sees is a complete structure and smooth logic, and it seems there is nothing that needs to be revised.

If the supervisor asks why these competitors were chosen, whether the conclusion would still hold with a different set of data, and what circumstances would make him abandon this recommendation, he may only then realize that he has not really thought it through.

The report can be submitted, but the judgment needed to produce the report has not yet been established.

What I mean by a working system is exactly these things. When receiving a vague task, knowing what to ask first; when reading contradictory information, knowing where to verify it; when several options all sound plausible, knowing what to base the trade-offs on. These standards also need constant revision, but one must first have a set of one's own.

In "The Secret of Caring for Life" in Zhuangzi, Cook Ding, when he first began carving up an ox, saw the whole ox before him. Three years later, he could already see its structure. Even after his skill became highly proficient, when he encountered places where sinews and bones crossed and the knife was difficult to apply, he would still become alert and slow down.

He even had judgment about when to slow down.

A very important part of professional experience is the same.

A report that others find smooth, you know that a certain page must be stopped at and checked again; a plan that others find reasonable, you know that one of its assumptions may have a problem.

This ability comes from encountering problems again and again, making choices, and then seeing the consequences of those choices. If the work process is always omitted and only the final answer is received, then many opportunities to build judgment are lost.

So, completing a piece of work and learning how to complete that work are two different things.

In January 2026, Anthropic published an experiment recruiting 52 mainly junior software engineers to learn an unfamiliar programming tool. In the follow-up test, the AI-assisted group scored 50% on average, while the handwritten-code group scored 67%; the largest gap was in the ability to find and diagnose errors.

Conclusion

I think interns will not exit history so quickly. But the way companies used to complete work while training newcomers will indeed be reconsidered.

In the past, the cost of training people was scattered throughout daily work. A supervisor revised a draft and explained a few more things; after a project ended, he took the newcomer to look at the actual results. The newcomer's experience accumulated in this process.

Both resumes can add a line, but after a few years they may lead to different places.

In the past, when looking for internships, people cared a lot about whether they could get into a famous company. In the future, they will have to ask one more question: after getting in, is there actually anyone willing to teach you?