Zuckerberg built himself an "AI boss"

Source: GeekPark

Written by: Huálín Wǔwáng

In 1972, Intel co-founder Andy Grove implemented a management method called “OKRs” within the company.

At that time, no one realized that this performance evaluation system, which looked like just a different form, would completely reshape Silicon Valley’s management philosophy over the next half-century.

Grove’s logic was simple— the faster information flows within an organization, the better the decisions, and the stronger the company.

Fifty years later, Mark Zuckerberg probably aims to do the same thing, but instead of a spreadsheet, he’s using an AI intelligence system.

The Wall Street Journal revealed this week that Zuckerberg is building a dedicated “CEO AI.” This system is still in development but already allows him to bypass traditional hierarchical reporting chains and quickly access various data and information from within the company.

It sounds like an advanced search box, but its underlying logic is far more complex.

In traditional large corporations, how many steps does information go through from the grassroots to the CEO? Department managers, VPs, SVPs, reports, meeting minutes… By the time the information reaches the decision-maker, it has either been filtered or become outdated.

Zuckerberg aims to solve this “information decay” problem.

This is not just a convenient tool but a surgical operation on the company’s power structure.

More notably, this CEO-exclusive AI system does not exist in isolation.

According to reports, Meta is now forming a complete AI tool ecosystem—employees are required to attend weekly AI training, participate in hackathons, and are encouraged to build their own AI tools.

Since February this year, Meta has become the first major tech company to formally include “AI usage” in employee performance evaluations. “AI-driven impact” has become a core assessment metric for every Meta employee.

In other words, Zuckerberg is not only building tools for himself but also reshaping the entire organization’s “operating system.”

01 From Metaverse to AI, the underlying logic of this gamble has changed

To understand Zuckerberg and Meta’s shift, we need to review the path Meta has taken over the past few years.

Everyone is familiar with the history of the Metaverse. In 2021, Meta changed its name from Facebook to Meta, announced a focus on virtual worlds, and then burned hundreds of billions of dollars over the next few years. However, users were not convinced, and the stock price plummeted. The problem with that gamble was that it was a “technology finding a scenario” story—building the palace first, then searching for willing residents.

This time, the AI strategy is entirely different.

Meta now bets on embedding AI directly into existing products and processes used by billions of people, not creating a new world but transforming operational machinery.

On March 16, Meta announced a partnership with infrastructure provider Nebius to purchase up to $12 billion in AI computing power by 2027. This year, AI investments are expected to reach between $115 billion and $135 billion, including large-scale collaborations with Nvidia and the construction of 30 data centers.

Meanwhile, over the past few months, Meta has acquired AI social media platform Moltbook and Singapore startup Manas AI, which focuses on personal AI assistants. The core technology of the latter aligns closely with the “CEO AI.”

This investment trajectory is very clear—first invest in computing power, then in scenarios, and finally use internal talent as the first testers.

02 Embedding AI into organizations is more dangerous than it seems

Of course, this path is not without pitfalls.

In mid-March, an internal security incident at Meta highlighted the risks of this “AI-first” culture.

An engineer used the AI system to break down a colleague’s technical problem on an internal message board. The AI, without any manual approval, published an answer on its own. Another employee then acted on this incorrect advice, leading to a leak of sensitive company and user data to unauthorized engineers. The vulnerability lasted nearly two hours before being discovered.

The danger of this incident lies not just in data leaks but in revealing a systemic vulnerability—once AI is deeply embedded into organizational workflows, its “out-of-control” behavior is no longer science fiction but a real engineering problem.

Analysts from MIT Sloan Management Review described this dilemma accurately—AI systems in organizations play dual roles as “tools” and “colleagues,” breaking traditional management boundaries.

When AI can collaborate, analyze, and even replace humans in decision-making processes, who is responsible? Who takes the blame if something goes wrong?

Zuckerberg’s use of the CEO AI to get faster information is not inherently problematic. But as the entire company accelerates down this path, the tension between “speed” and “safety” will intensify.

03 When CEOs manage themselves with AI, what are employees thinking?

There is an even more subtle aspect worth discussing.

Meta laid off about 11,000 employees in 2022 and another 10,000 in 2023. Now, performance evaluations are linked to AI usage, with management repeatedly emphasizing “flattened teams” and “enhanced individual contributions.”

Employees are not fools. They understand what these words imply.

If an AI system can help the CEO bypass hierarchy and access information directly, do middle managers responsible for “information transmission” still need to exist? If every employee’s productivity can be multiplied with AI tools, what percentage of the current workforce is actually needed?

Sam Altman recently said at the India AI Summit, “AI can be a better CEO for a large company than any human.” Sundar Pichai also publicly stated that AI might “replace him” within a year.

These statements sound humble, but in the context of Meta’s “AI-driven performance” approach, they are more like warnings.

Tech CEOs are signaling through action: AI is not just an efficiency tool but the starting point for organizational restructuring.

What this means for employees remains uncertain. But one thing is clear—by 2026, over 70% of the Fortune 2000 companies will have moved AI systems from pilot projects to full deployment.

The window for hesitation is closing.

Grove invented OKRs because he realized that an organization’s most scarce resource is not money but attention. Zuckerberg’s creation of the CEO AI follows the same logic— in a company with 70,000 employees handling billions of pieces of content daily, the speed at which information reaches decision-makers determines how fast the company can run.

But this time, the costs and boundaries of acceleration have yet to be fully understood.