The profile of talent that enterprises seek in the AI era has undergone a transformation.

Baidu’s sweeping reforms to its job level system have become the most talked-about topic in today’s internet technology sector.

The company has fully integrated professional and managerial career paths below the middle management level, replacing the long-standing letter-based job levels (such as T, P, E, Band, and M) with a numerical system ranging from Level 5 to Level 12, effective May 1st. The rationale behind this change is clearly stated in the announcement: “The AI era demands interdisciplinary talent who possess both professional expertise and leadership skills, capable of taking charge independently.”

Looking at the 2025 data, Baidu reported total revenue of 129.1 billion yuan, marking a 3% year-on-year decline, with net profit experiencing a significant drop. The stock price has corrected by approximately 26% from its early-year high, with market capitalization dipping from nearly 440 billion Hong Kong dollars to 328.6 billion Hong Kong dollars at one point.

Despite AI business revenue accounting for 43% in the fourth quarter and reaching 40 billion yuan for the full year, the rapid growth of AI has yet to offset the decline in traditional advertising revenue.

Faced with pressure, companies either retreat or adapt. Baidu chose the latter.

However, this article is not about recounting Baidu’s struggles. When viewed within the broader industry context from 2025 to 2026, a clearer picture emerges: China’s internet giants are collectively “rebooting” their job level systems.

01

What Barriers is Baidu Breaking Down?

To grasp the significance of Baidu’s reforms, we must first identify what barriers it is dismantling.

In Baidu’s former system, T represented the technical sequence, P the product sequence, E the engineering sequence, Band functional support roles, and M the management sequence. These operated as five parallel career paths. A T8 technical expert and an M-sequence manager followed entirely different promotion tracks.

This dual-track or “multi-track” design once addressed a core issue: it allowed technical specialists who preferred not to manage people to receive recognition and compensation without needing to “climb the corporate ladder.”

However, the cost became apparent over time. Mechanisms for switching between tracks often remained theoretical.

A product manager in the P sequence wanting to transition to management, or a technical expert aspiring to lead a team, found the process mentally and physically exhausting. Eventually, these tracks became invisible barriers. More critically, “hybrid talents”—those with both technical acumen and business sense—struggled to find their place in this system.

Baidu’s solution is straightforward: dismantle the barriers entirely. Rather than adding another track or opening a door, all career paths below the middle management level are unified into a single numerical ladder ranging from Level 5 to Level 12. Existing job levels are mapped to the new system through a “one-to-one correspondence” to ensure fairness during the transition.

A product manager in the P sequence and a manager in the M sequence might end up at the same numeric level after mapping. The “origin” of their track no longer serves as a distinction.

At the same time, Baidu has set two criteria for talent selection: a young mindset and the ability to think using first principles, explicitly stating its intention to provide “opportunities for young people to take on significant responsibilities.”

In March of this year, Baidu launched a summer internship program for the 2027 graduating class, offering over 5,000 internships, with more than 90% of the positions related to AI, covering cutting-edge fields such as large model algorithms, multimodality, and autonomous driving. This marks the largest-scale summer internship recruitment in the company’s history.

New talent is pouring in, old barriers are being dismantled, and both forces are pulling in the same direction.

In fact, Baidu is not the first company to overhaul its job level system. It could even be said that Baidu is relatively late to the “barrier removal” game.

Tencent simplified its original complex 6-level, 18-grade system into a professional job level system ranging from Level 4 to Level 17. A few years ago, it also made a detailed adjustment by amortizing the year-end service award (i.e., the 13th-month salary) over 12 months and incorporating it into the base monthly salary, with housing subsidies also integrated into the monthly pay. Around the same time, Ant Group also split the 13th-month salary for employees at Level 17 and below into their monthly wages, while for employees at Level 18 and above, the 13th-month salary was incorporated into their year-end bonus and linked to performance.

On the surface, these appear to be salary structure adjustments, but the deeper intention is to make the correspondence between salary and value more direct.

Alibaba initiated the “de-P-ization” reform in 2023, starting with Taobao and Tmall, replacing the once-glamorous P sequence with a numeric system ranging from Level 14 to Level 28. By 2025, Alibaba had fully opened up internal network permissions, removing barriers to cross-group transfers between business units.

ByteDance made the most aggressive move. At the end of 2025, it announced a complete switch from the original “5-level, 10-grade” competency system to a numeric job level system ranging from L1 to L10. Subsequently, CEO Liang Rubo declared three words at an all-hands meeting: “Reach for the Peak.” Salary adjustments increased by 1.5 times compared to the previous cycle, with job level entry points and ceilings raised across the board, and global salaries aligned with top-tier levels.

Maimai data shows that in 2025, 17.44% of employees in enterprises with over 10,000 employees received salary increases, while intern positions at ByteDance could command monthly salaries of 10,000 yuan or more. Liang Rubo is taking an even bolder approach: using profit advantages to “make poaching talent an expensive proposition.”

A uniform trend of “de-labeling,” numeric job levels, and track integration has emerged. This is no longer a matter of internal management preferences for individual companies; the entire industry is solving the same strategic challenge.

02

What is Driving This Reform?

The variables in this strategic challenge are few, arguably only one: AI is no longer just a slogan.

In 2025, Baidu’s AI business revenue reached 40 billion yuan, with intelligent cloud revenue increasing by 34% year-on-year; AI-native marketing service revenue grew by 301% year-on-year; and AI application revenue exceeded 10 billion yuan for the full year.

In the job market, the average monthly salary for newly posted AI algorithm engineer positions increased by over 36% in 2025. In the spring recruitment season of 2026, the number of AI positions increased by approximately 12 times year-on-year, with the average monthly salary reaching 60,738 yuan.

These numbers, translated into plain language, mean: the AI business is surging, but there are far from enough people capable of driving it.

Many analyst reports on talent trends have made it clear: the profile of talent that enterprises seek in the AI era has undergone a transformation. The foundational layer demands a comprehensive AI transformation of traditional general R&D roles, requiring both AI technical expertise and a deep understanding of vertical industries; the executive level requires AI leaders to possess a “technology + business + implementation” trifecta of full-chain control capabilities.

During our visits to the job market, we also found a significant increase in demand for roles related to AI agents, with companies favoring interdisciplinary talents who possess AI application capabilities and can solve practical problems in specific industry scenarios.

This is precisely the blind spot of the traditional dual-track system. Under the old framework, a person was either labeled as a “professional technical talent” or a “management talent.” The system assumed these two paths were parallel and mutually exclusive.

However, the logic of product development in the AI era has changed. Large model training is not merely an engineering problem; it involves product definition, scenario selection, user feedback, and commercial closure. A technical expert who only understands code but not business, or a team leader who only understands management but not algorithms, cannot succeed in AI product development alone. What you need is someone who can simultaneously consider technical feasibility and commercial logic.

Baidu’s statement in the announcement—“possessing both professional expertise and leadership skills, capable of taking charge independently”—essentially represents a public farewell to the old division of labor logic.

Of course, a technical question worth pursuing is: Why have major companies, as if by prior agreement, chosen “pure numeric labels”?

The labels in the old system—T, P, M, Band—carried far more than just job level distinctions. They represented a whole set of identity affiliations. The P sequence was once synonymous with “technical aristocracy” within Alibaba, while the T sequence at Baidu symbolized hardcore technical prowess. These letter labels created an invisible sense of hierarchy among employees and generated unnecessary friction in cross-departmental collaboration.

Internet giants previously defined job levels based on ability tiers, essentially “determining the shape of the pit based on the size of the carrot,” which is logically flawed. Job levels should be based on the responsibilities, authority, and value of the position, while ability is an evaluation of the individual.

The confusion between the two led to a “shortage of monks and an excess of incense,” where an increase in ability evaluations demanded corresponding management positions, but management roles are limited. The gap between expectations and reality created numerous conflicts.

Switching to a string of abstract numbers does not automatically eliminate these issues, but it at least formally flattens the hierarchical distinctions of identity attributes. When job levels change from “T8” or “M4” to “Level 10” or “Level 11,” cross-track mobility no longer requires any additional explanations or approvals. A single number serves as a unified standard.

In this sense, numeric labels represent an organizational language with the lowest institutional costs.

03

Can Baidu Turn the Tide by Breaking Down Barriers?

Any organizational transformation must ultimately answer a simple question: Can it alter the real dilemmas faced by the company?

Baidu’s full-year revenue decline in 2025 was due to both macro-level contraction in the advertising market and long-term structural issues eroding its search business market share. The 43% AI revenue share in the fourth quarter was a bright spot, but it also highlighted the severity of the contraction in non-AI businesses.

Breaking down barriers does not solve business problems per se. However, it may produce chain reactions at three levels:

First is internal mobility. After the tracks are integrated, a person with a solid technical background and business ideas can more smoothly take the lead on projects. This reuse of “multi-faceted” capabilities is the organizational form most needed for AI businesses.

Second is external attractiveness. As job level labels across the industry move toward standardization, Baidu’s unified 5-12 level numeric system makes salary grading negotiations for external talent more transparent. A person transitioning from ByteDance’s L sequence can easily see the corresponding relationship.

Finally, this directly relates to the gateway of salary flexibility. During the dual-track period, each track had its own independent salary bandwidth. After track consolidation, the salary bandwidth theoretically merges as well. This change reserves wider operational space for Baidu in the AI talent war. Facing a candidate with both technical and management potential, HR is no longer constrained by track bandwidth when determining the salary ceiling.

Looking beyond, this collective job level system transformation is fundamentally driven by a shared anxiety: AI talent is extremely scarce.

In the spring recruitment season of 2026, the number of AI positions surged by 12 times year-on-year, with the average monthly salary for newly posted positions exceeding 48,000 yuan. ByteDance directly raised the salary ceiling, with a clear calculation: using talent inflation to build barriers and make the cost of poaching by competitors prohibitively high.

Tencent’s precise poaching of talent from ByteDance for its large models, such as the public secret of Feng Jiashi, former head of visual foundational research for ByteDance’s Seed large model team, joining Tencent, shows that all parties are voting with real money on a judgment: future internet competition has shifted from vying for traffic and algorithms to competing on “intelligence density.”

Rebuilding the job level system from scratch, while seemingly a task for the HR department, is actually about making each company’s organizational framework more suitable for accommodating interdisciplinary talents and establishing a more competitive pricing mechanism in the vast talent marketplace.

Baidu’s chosen path has a unique feature: it is the latest but most decisive among the major players. Instead of adding a “interdisciplinary track,” it directly merged all tracks below the middle management level into one. This represents a rather thorough “reboot” reform.