The AI-Driven Reshuffle of the Talent Market: Who Decides Your Worth?

How humans coexist with AI.

Global tech layoffs continue. Since February this year, multiple companies have announced job cuts, some of which are the largest in their history.

Oracle carried out an overnight "cross-border mass layoff," with an estimated 30,000 employees affected. Block, the fintech company founded by former Twitter CEO Jack Dorsey, slashed 40% of its staff in one go. Meta is also reportedly planning a new round of major layoffs — nearly 20% (about 16,000 employees) may lose their jobs.

Massive layoffs in the tech industry are spreading worldwide. On April 8, Nikkei Asia reported that in the first quarter of this year, global tech companies laid off 80,000 people, with nearly half of these reductions directly or indirectly attributed to AI implementation and workflow automation.

These figures seem to confirm the most widespread AI anxiety — people constantly worry that their jobs will be replaced by AI, and this is actually happening. Dorsey did not hide the fact: the layoffs are not because the company is in trouble, but because "AI is fundamentally changing the model of building and operating a company."

Against this backdrop, not only the tech industry but the entire market's definition and valuation of talent are shifting. On this watershed redrawn by technology, some are passively eliminated, while others actively restructure their positions.

01 AI proficiency becomes a job requirement

Grace is a programmer at an overseas tech company. AI has completely transformed her way of working.

Now she barely writes code manually — she only needs to input a prompt, and AI generates and modifies the code. Even deleting a single line is done through AI. The advantage is that AI understands why the change is needed from the context, whereas manual changes would be treated as a new input, disconnected from other code.

In Grace's team, a colleague once used Claude Code to generate most of the code, made manual modifications, and then submitted manually. However, that could not be counted as a Claude Code submission. The manager approached this colleague and stressed that Claude Code must be used for submissions, otherwise it would affect the team's AI usage rate.

Actually, a few years ago, her company started encouraging employees to use Copilot and Windsurf. These early AI coding tools rely on large code libraries to quickly predict and write complete code based on programmer input, and they can also handle submissions.

Claude Code is completely different — it is directly integrated into the system with higher permissions. It can read files, delete/modify files, and is much smarter.

As a result, the company's attitude has changed significantly. Copilot and Windsurf were "optional." Claude Code, however, has been mandatory since January this year.

AI usage rate has become one of the core metrics for management to measure programmer workload. According to Grace, in addition to basic code volume, many tech companies now track granular metrics such as which tools are used, how many lines each tool submits, and the proportion of AI usage per employee.

Not using AI tools hurts productivity. Although coding speed and output have not yet become rigid KPIs, the average is rising. Grace says it is not an explicit rule, but "if you don't keep up with others' speed, you'll fall behind, right?"

Domestic internet companies are experiencing similar shocks almost simultaneously.

After 2026, Liu Yun's software company also mandated the use of AI development tools such as Claude Code for all employees. At the same time, external experts were invited to share success stories, claiming that a single person using AI could complete the workload of a senior programmer (two years' worth) in just two months.

Everyone feels a strong "AI push," and everyone knows deep down that the company's aggressive promotion of AI across business processes is essentially "how to replace yourself with AI": programmers use AI to auto-write code, testers use fully automated AI testing. Once these tasks become fully AI-driven, everyone knows what will happen next, but they feel helpless and can only go with the flow.

Programmers are undoubtedly among the first to feel the impact of AI. But across all industries, AI is hitting different jobs to varying degrees — both in terms of capability definition and compensation.

Lin Cheng, a teacher at a middle school in Guangdong, is also facing the pressure of the AI era. Since DeepSeek became popular, every public class or competition now includes the words "AI empowerment," and her school has become an AI pilot school.

Lin Xue has long served the content dissemination of a multinational retail brand. Previously, design and text deliverables followed market rates. After AI design tools like Jimeng and Keling emerged, the client's payment logic changed: the cost of a video or poster now consists of tool membership + token consumption + half a day of labor cost. The total cost dropped by nearly two-thirds compared to before.

Lin Xue says this is a disaster for designers — a severe erosion of their value. But she also understands that this trend is irreversible.

Looking across the workplace, the trend of "everything AI" is even more apparent.

Yao Jinbo, chairman of 58.com, revealed in an interview with China Entrepreneur that although the company retains its old evaluation system, when promoting cadres or adjusting organizational structures, the most important factor is AI capability: "If I judge that a person does not understand AI or lacks future vision, I will never put them in a management position."

This mindset is spreading top-down throughout the job market. Maimai's "Insight Report" shows that employer requirements for candidates' AI capabilities are rising rapidly: over 34% of new job postings explicitly mention keywords like "AI" or "large model." Recruiter Xiao Chai also noted that "the ability to use AI products" has become a must-ask question in recruitment — "there was almost no such requirement before."

The Insight Report also indicates that nearly 80% of surveyed employees work at companies that have introduced AI-related requirements, with over 30% of companies implementing matching assessment or training mechanisms.

Thus, AI skills are transforming from a plus to a hard threshold. However, according to Xiao Chai, for now, while AI proficiency helps in job hunting, it does not necessarily lead to higher pay — "unless it's for AI programming roles."

But how exactly should one judge whether a person has AI capabilities?

As a long-time internet professional, Liu Yun says that in the past, candidates for programmer roles typically had to practice algorithm problems or even hand-write code on the spot to prove their real ability. But now AI can easily solve algorithm questions and the theoretical "eight-part essay" that used to be part of the interview. How to identify talent has become a new challenge for companies.

"But our company is only reducing headcount, not adding, so I don't know if job requirements or interview processes have changed," Liu Yun told Xin Mei.

02 The first job for young people may disappear

Under the onslaught of AI, which job types will shrink? And which will be completely replaced? These have been hot topics since AI emerged.

Anthropic's March "Economic Index" shows that jobs centered on information processing — customer service, market analysis, design — are most impacted by AI: because AI now performs tasks that previously required advanced degrees, the professional moats of these roles are collapsing.

These are not just cold research data. At the company where AI entrepreneur Tao Zhe previously worked, the impact of AI has precisely corresponded to specific roles.

He used to oversee AI-related business, and one core project was introducing AI into online customer service and sales. This initiative generated tens of millions in annual revenue for the company, but also led to the reduction of some sales positions.

"What companies need are good salespeople, who are already hard to find," Tao Zhe explains further. "Whether in tech or sales, senior talent is always in short supply, and that's just the market. Now that AI is replacing mid-to-low-level positions, those roles naturally shrink. But when I want to hire senior talent, the supply in the market hasn't increased, so they are not currently at risk of being replaced."

The same logic applies to programmers. According to Tao Zhe, after some domestic tech companies aggressively pushed employees to use AI, they did adjust and reduce headcount — mainly "business development programmers."

He explains: "These programmers' work is usually more about business implementation and requirement delivery, with stronger repetitive elements and relatively lower requirements for underlying system capabilities." Senior programmers, on the other hand, have the ability to "cover the bottom line" — they build the framework, and others implement functions based on the framework's interfaces. "Even if application-level functions break, the framework usually has built-in degradation, fault tolerance, and exception handling mechanisms to prevent overall service loss — these capabilities come from their long-term engineering experience."

In fact, many junior programmers are not just "worried about being replaced" — they "have already been replaced." But Tao Zhe finds that senior programmers rarely have this concern; some are even job-hopping. "Because when you reach a senior role, what you compete on is often not a specific skill, but a comprehensive ability across business, technology, and organizational coordination."

As AI replaces a large number of low-level jobs, employer requirements for experience have changed significantly.

The Insight Report shows that in the 2026 spring recruitment season, positions requiring more than three years of experience account for over 70%, with demand for the 3-5 year and 5-10 year experience segments growing notably. In contrast, positions for candidates with less than one year of experience decreased by about 20% year-over-year, showing a clear "de-juniorization" trend in the market.

"AI affects salaries at large model companies, but for ordinary companies, changes are not huge," says Xiao Chai. The report mentions that AI talent competition dominates spring recruitment, with job postings soaring 12 times and average monthly salary reaching 60,738 RMB.

However, regarding junior positions, Xiao Chai admits that many internships have indeed been replaced by AI, but recruitment hasn't stopped completely.

Grace also mentions that interviews have become much harder, and the bar for newcomers is higher. As for salaries — neither a universal drop nor rise — "it's just not very dynamic." The reason is not hard to understand: "If one engineer plus $1,000 in tokens can do the work of three of you, why would your team need so many people? If your team's output doesn't increase, why keep so many people?"

So Tao Zhe fully understands the severe AI anxiety among new college graduates.

"Because fresh graduates, whether as programmers, customer service reps, or salespeople, have lost the path from junior to senior. No one will have a reason to hire junior talent — even at a very low salary, you can't compete with AI. In the current environment, only mid-to-senior talent complements organizations."

This means many young professionals are losing opportunities to gain experience.

He Fan, dean of the China Development Research Institute at Shanghai Jiao Tong University, said in an interview that the AI wave is causing many young people's first jobs to disappear — and in the future, children may have to pay to get their first job.

03 In the AI era, taste and judgment are scarce

Today, Claude Code has become an indispensable coding assistant for Grace, but its limitations are also obvious. Precisely because she recognized these early on, Grace — at the "front line" of AI — is far less anxious than outsiders might imagine.

She points out that all open-source code is training data for Claude, so it's very familiar with that. But complex internal company code is new knowledge to it. Moreover, Claude has a 100,000 token limit; beyond that, it compresses, and sometimes becomes a bit "dumb."

"When the human brain learns something, you sleep on it, and after compression you know what's important and what's not. What we forget is the unimportant stuff. But when Claude compresses, it might forget some very important things," Grace explains. "Every company has complex internal systems. People who jump from company to company bring understanding of previous systems, but Claude can't do that."

So her habit is to open many windows and work on several projects simultaneously. For each project, she first discusses a rough plan with AI, then handles the details step by step. "AI has stronger computing power, but its structure is different from the human brain. So if humans help break it down, it's definitely the most efficient method."

However, even after humans make necessary decisions and breakdowns, AI still has flaws. So while it generates a large amount of output, it also creates a large number of problems that need human resolution.

Simply put, "entropy has increased." Grace says, "Before, when you built something, you built it slowly. Now you build so much so fast, testing can't cover everything, so things break here and there. We actually have many more problems than before, and everyone is very tired. So we still need many people to solve the new problems derived from AI."

Liu Yun feels the same. Although everyone has accepted the reality that "AI will replace programmers," programmer positions still exist in companies for now. Perhaps some organizations will rename the role, like "AI engineer."

But that doesn't mean traditional accumulated experience and skills are discarded or useless. Liu Yun says, when using AI to write code, someone who knows how to code will definitely have an advantage over someone who doesn't. To solve problems with AI, the premise is to understand what problem you are facing. At this point, taste and judgment become more important — and scarcer.

If AI has already turned the tech industry upside down, some traditional industries are still in the "loud thunder, small raindrops" stage.

At Lin Cheng's middle school, how exactly should AI empowerment be done? The school provided no guidance.

So, with this question, she took the opportunity of public classes to visit schools in other provinces, only to find that form often outweighs substance — sometimes even "technology for technology's sake." For example, some schools require every student to have a tablet, while others add a few "digital human" clips to presentation materials.

"The so-called AI empowerment only appears in public classes, because that's the only opportunity to showcase it," Lin Cheng says. "It feels like a hot topic that everyone wants to chase..."

So for now, "AI empowerment" is not yet a mandatory task. As for how to actually apply AI in teaching, she finds it difficult at this stage; older teachers may even resist somewhat.

However, five ministries including the Ministry of Education recently issued the "AI+ Education Action Plan," encouraging AI interdisciplinary teaching in basic education, integrating AI education into after-school services and research-based learning, and including AI in teacher qualification exams and certification. The integration of AI with daily teaching is foreseeable.

The consensus among several interviewees is that AI and humans will gradually integrate, rather than 100% replacement.

"It's not as you imagine — using AI directly to 'eliminate' people. It will restructure everyone's work. For example, AI sales will spawn a new role: AI sales strategy. This role will help define which parts can be handled by AI, which must involve humans, and which need human guidance for AI iteration before being used in industrial settings," Tao Zhe says. Xiao Chai also believes that the "AI helps employees improve efficiency" working model will become mainstream.

As AI continuously raises the floor of efficiency, what becomes scarce is no longer the ability to complete tasks, but the ability to judge, break down, and bear uncertainty. Perhaps more important than "whether you will be replaced" is the extent to which you can actively choose how you are needed.

(Per interviewee requests, names Grace, Liu Yun, Lin Xue, Lin Cheng, and Xiao Chai are pseudonyms.)