2026 Most Brutal Workplace Shift: It’s Not AI Stealing Jobs – 36氪

This post was originally published on this site.

Not long ago, I chatted with a former colleague who worked as an HR at a top internet company, and we talked about this year’s campus recruitment. He said something that made me feel quite upset:

“We are no longer recruiting new graduates now, we directly hire people who can take on real responsibilities.” 

I asked him, then who will do those jobs that were originally used to let new people practice? 

He smiled and said, AI does it. 

At that moment, I suddenly realized that the question “Will AI replace me” that many people have been anxious about for a whole year was actually asked in the wrong direction. 

AI is not taking away your job, it is taking away the ladder for you to get promoted. 

What I want to talk about today is called “the disappearing middle tier”. 

The first to disappear are the least valuable tasks

In October this year, the General Office of the State Council issued a document named “Opinions on Adapting to the Development of Artificial Intelligence Technology and Helping Build an Employment-Friendly Development Mode”, which covers 18 tasks in six aspects. The first sentence at the beginning is “Employment is the most fundamental livelihood of the people”. 

The fact that a national document puts this sentence at the top itself shows that the problem has reached a level that requires a systematic response. 

ShanghaiTech University has conducted a research, whose data base covers a total of 743 million recruitment information from mainstream recruitment platforms from 2022 to July 2026. 

The conclusion is: basic operation, financial accounting, graphic and text content, telesales the demand reduction in these four fields is the most obvious. The contraction effect of AI is concentrated on entry-to-mid level people with monthly salary of 5000 to 8000 yuan and 1 to 3 years of work experience. 

Put it in plain language: In the past, positions that required new employees to sort out documents, input data, and write standardized communication scripts in a company are no longer set up now. 

I can fully understand where the harm of this change lies. 

What it cuts off is not a single position, but a person’s “practice period”. 

Looking back on the time when I just graduated, my first job was doing the most basic tasks: sorting out materials, writing meeting minutes, making spreadsheets for leaders, finding materials, and assisting in completing reports. At that time, I thought these tasks were trivial and worthless, but looking back now, it is precisely these tasks that let me figure out how a company operates, and gave me the opportunity to gradually transform my student mindset into a professional workplace mindset. 

But now, this period of time that “allows you to be clumsy” is being taken away from many industries. 

It’s like when you first entered the workplace in the past, you could say “I don’t know how to do it but I can learn”. 

Nowadays, there is no time and opportunity left for you to learn. 

What is more worthy of vigilance is another set of data: 

The programming industry has shown an obvious pyramid effect: the number of junior developers is decreasing, the number of mid-level developers is slightly decreasing, while the number of senior developers is not decreasing but increasing. 

That means this industry is not stopping recruiting people, it no longer needs people who are “still learning”. 

On the other side, middle management positions are also being cut

If only entry-level positions disappear, it is only a problem for new graduates. 

But this year’s more drastic change has taken place in the middle tier. 

On November 16, Tencent’s headquarters made public the internal conversation notes, clearly proposing to accelerate organizational flattening, take small teams as combat units, and also said a very straightforward sentence: 

Do not let cadres who only pass on messages, do not want to make progress and occupy resources become obstacles.

This statement sounds harsh, but it points to a real problem. 

The larger the scale of a company, the more levels it has. The boss cannot manage hundreds of people directly, so they can only set up supervisors, managers and directors level by level. But in actual operation, many middle-tier people gradually become a link only responsible for collecting and conveying information: 

Front-line staff fill in forms, supervisors summarize, middle managers rewrite them into presentation materials and then hand them over to senior management.

So you will see a very absurd phenomenon: 

There are only five or six front-line employees who actually do the work, but there are two or three supervisors and managers who keep pressing these few people for progress every day. 

This kind of “information transfer” work is exactly what large language models are best at. 

An e-commerce company in Hangzhou directly canceled the 8-person PMO team and the 6-person data report team, and cut the 12-person test team to only 3 people, saving 4 million yuan in labor costs a year. 

So you see, the two ends of this round of changes are happening at the same time: 

Entry-level positions at the bottom are disappearing, management positions in the middle are being compressed, and caught between the two, the most uncomfortable are exactly those people who have worked for three to five years and are planning to move up.

It’s not that you don’t work hard, it’s that the path has become narrower. 

But on the other hand, people with rich experience are becoming more valuable

Up to this point, it seems that all the news is bad. But in this year, there is another set of completely opposite data. 

Have you paid attention to this year’s national civil service examination? For the first time, the age limit for application was relaxed from 35 years old to 38 years old, and the age limit for fresh master’s and doctoral graduates was relaxed from 40 years old to 43 years old. 

This is the first loosening since the “under 35 years old” rule was established in 1994, and the civil service joint examination of more than 20 provinces has followed up. 

At the same time, there is an obvious “de-juniorization” trend in the recruitment market: 

Positions requiring more than 3 years of experience account for more than 70%, and positions that only recruit new people with less than 1 year of experience have decreased by about 20% year on year. 

There is a detail that I am particularly impressed with. 

For the solution position of a certain cloud vendor, the recruitment requirement has changed from “recruiting new people with less than 3 years of experience to sort out documents” to “requiring more than 5 years of experience, able to judge risks and take responsibility for projects”. 

Experience, for the first time, has been revalued from a “cost” to an “asset”.

I always remember a sentence a former leader said to me: Doing work is like forging iron, punting a boat and grinding tofu, the smaller the matter, the better you need to do it. At that time, I just entered the workplace, and I always thought this sentence was a bit old-fashioned and rigid. 

Now I have a new understanding: 

Doing small things well is not to prove that you can bear hardships, but to let you accumulate judgment in repetition. Because AI can finish the work for you, but it cannot replace you to judge whether the work should be done in this way.

A chief macroeconomist at the International Labour Organization put forward a term called “verification tax”. 

It means that AI will produce hallucinations, and the work it finishes still needs to be checked by humans. 

Therefore, what is really replaced is never the entire position, but those links that “do not need human verification”. 

This also explains why senior positions are increasing rather than decreasing: 

The stronger AI is, the more valuable the person who can sign, take overall responsibility, and say “this won’t work” will be. 

Final note:

I have read too many articles about AI this year, and most of them are asking the same question: 

Will AI replace me? 

But I think the real change this year is not replacement, but re-pricing.

Some people are re-priced as cost, while others are re-priced as asset. The difference does not lie in who works harder, but in who holds the things that “AI cannot do and must be taken responsibility for by humans”. 

So instead of worrying every day whether you will be laid off, it is better to ask yourself a serious question: 

If the company puts an AI tool on my desk tomorrow, how many of the tasks in my hand can I judge, be responsible for, and take overall charge of? 

This number is roughly the real safety line for you in this company. 

After all, AI can do the work for you, but it cannot take responsibility and make judgments for you. 

And whether you are valuable, in the final analysis, depends on how important the things you can take responsibility for are. 

This article is from the WeChat official account “Ray Xian Sen” (ID: RayZhang1993), written by Ray Xian Sen, and published with authorization from 36Kr.

Leave a Reply

Your email address will not be published. Required fields are marked *