In January, Mark Zuckerberg and Meta executives gathered for their annual retreat at his estate in Hawaii. There, they discussed a radical overhaul of how the company would operate in the age of artificial intelligence.

As noted by Reuters

The plan, code-named “Project Organizational Transformation,” envisioned turning Meta into a company where artificial intelligence would perform a significant share of the daily work of thousands of employees. Smaller, highly specialized teams of people would oversee virtual workers.

While modeling different scenarios, executives considered reducing the size of many teams by as much as 60%. Some employees might be offered transfers to new divisions, while the rest would be laid off. One human resources executive suggested that the scale of the cuts could be no smaller than in the previous round, when Meta reduced its workforce by roughly 25%.

The restructuring was planned in two waves: the first in May and the second in November. In addition to layoffs, the plan included closing vacant positions and dismissing employees the company considered underperformers.

However, on the evening of May 19, just hours before the first wave was due to begin, Zuckerberg changed course. The next day, Meta cut 10% of its workforce but abandoned the phase planned for November.

By then, employees had already begun openly protesting the changes. They believed Meta’s artificial intelligence initiatives were partly intended to replace people. Internal metrics also suggested that the autonomous agents on which the company had pinned its hopes were not delivering the expected productivity gains.

Meta confirmed the project’s existence, describing it as a yearlong program to reduce costs, restructure teams, and move employees into priority areas, including the creation of training data for artificial intelligence models.

The company acknowledged that it had considered scenarios involving cuts of up to 60% to individual teams, but emphasized that it had not planned to lay off 60% of its entire workforce. According to Meta, the scenarios included both layoffs and transfers to other roles, and the second wave was canceled before its scope had been finalized.

How Meta Tried to Become an “AI-Native Company”

After ChatGPT launched in late 2022, Silicon Valley executives began actively discussing how generative artificial intelligence would change the future of work. Autonomous agents attracted particular interest because they could do more than answer questions: they could independently take actions such as buying products, booking trips, or building applications.

Meta executives embraced the idea that companies should not simply add artificial intelligence to existing processes but completely redesign their work around it. Internal documents described tools and agents that would interact with one another, automated workflows, and product development based on an “AI-first” principle.

Alex Schultz, Meta’s chief data officer, and product executive Naomi Gleit joined the discussions. During trips to Asia, they studied startups that had built their organizational structures around artificial intelligence from the outset. Meta also conducted its own research and launched pilot projects.

Gleit said she spent much of last year at Meta’s Singapore office. According to her, local approaches inspired teams in California and New York, while some ideas originated within the company itself.

One of the first experiments was launched last year by Ime Archibong. His team created five small technical pods, each staffed by two or three engineers and a designer. They were expected to use artificial intelligence tools and abandon traditional six-month planning cycles, building prototypes in four-week sprints.

In basketball, a fast break lets you take more shots – and better ones. We expect the same effect from artificial intelligence tools: they allow us to explore more ideas at lower cost and with greater precision.

– Ime Archibong

In October, Archibong’s team presented a broader guide for transitioning to the new model. Traditional product designer and engineer roles were supposed to disappear, while members of the technical pods would share the title “builders.” Middle management was slated for cuts, and daily priorities were to be determined through agent-assisted analysis.

At the beginning of the year, Zuckerberg launched the organizational transformation project and instructed executives to change the management structure. By June, at least 11 divisions, including engineering and research teams, had already introduced small working pods.

One division called the model the “village approach.” Leaders of major areas, known as organization heads, were supposed to make decisions about performance reviews and promotions with support from human resources and unspecified “artificial intelligence systems.” Each organization head was responsible for 30 to 50 employees, while pod leaders coordinated daily work without formal management authority.

One employee assigned to lead such a pod wrote on an internal message board that he was not receiving management training and did not have access to employee evaluation tools. Meta responded that decisions about performance reviews and promotions were made by people, not artificial intelligence.

Around the same time, the company launched a tool to identify “irreplaceable talent.” Documents referred to a hypothetical “10x employee” – an exceptionally strong engineer who could supposedly do the work of an entire team. Meta planned to use some of the savings from the cuts to offer large compensation packages to attract and retain such specialists, particularly artificial intelligence engineers.

Employee Discontent and Productivity Problems

In March, reports emerged of possible cuts that could affect 20% or more of Meta’s employees. This would not have been the company’s first major workforce review: in late 2022 and early 2023, it cut roughly a quarter of its staff.

The news alarmed employees. Management did not openly discuss the situation with rank-and-file workers, while managers were advised to tell their teams that their roles were “evolving” under the influence of artificial intelligence.

In April, Meta confirmed plans to cut about 10% of its workforce on May 20. Zuckerberg attributed the move to substantial capital expenditures. At the same time, the company was transferring engineers to a new applied artificial intelligence engineering division, where they created programming tasks to train models.

Some employees described the work as monotonous. By the end of May, the size of some engineering teams had fallen by 30% after transfers and layoffs.

Meta also installed software on employees’ devices in the United States that recorded keystrokes and mouse movements. The data was intended to help train agents to reproduce how people interact with computers. Employees feared they were effectively training a system that would eventually replace them.

On the internal Workplace platform, employees posted angry messages and sarcastic jokes. In replies to executives, they shared images of elephants, implying that the cuts remained “the elephant in the room.” Some employees engaged in arguments with Meta CTO Andrew Bosworth, who was responsible for the artificial-intelligence-based transformation.

One employee even sarcastically compared Zuckerberg’s internal initiative for small businesses to Prometheus, who, according to ancient Greek mythology, gave fire to humanity.

Morale inside the company deteriorated sharply: in Meta’s semiannual survey, the share of positive responses fell from 74% to 55%. At the same time, employees stepped up efforts to organize in defense of their labor rights.

Internal data also cast doubt on the benefits of the new technology. The volume of code produced with artificial intelligence rose sharply, but this did not lead to a proportional increase in productivity. According to a post by Bosworth, changes to internal software platforms and infrastructure increased by 220% over the year, while the number of new or improved features available to users rose by only 36%.

Infrastructure teams warned of reliability problems linked to the rapid growth of AI-generated code. Another post said that out-of-control agents had carried out large-scale destructive actions that people would be unlikely to take. The number of major technical and security incidents, including service outages and potential data leaks, rose by 40%, while the time needed to resolve them increased by 70%.

In early June, hackers exploited a vulnerability in Meta’s new customer-support chatbot and gained access to notable Instagram accounts, including the inactive White House page from Barack Obama’s presidency.

Under pressure from employees, technical problems, and investors’ scrutiny, Zuckerberg changed course. The second restructuring wave, scheduled for November, was canceled, but the 10% workforce reduction on May 20 still went ahead.

After layoff notices were sent, Zuckerberg wrote to employees that he did not expect further companywide cuts through the end of the year and wanted to provide people with greater stability.

Management tried to restore morale: it suspended the program tracking mouse movements and keystrokes, allowed some engineers to return to their previous teams, and tasked CFO Susan Li with working to improve conditions for employees.

At an internal meeting in July, Zuckerberg acknowledged that he had misjudged the pace of the reorganization. He said autonomous-agent technology was developing more slowly than he had expected, although it was supposed to demonstrate greater value over the following three to six months.

From “Reducing Work” to “Betting on People”

After softening the most radical elements of the transformation, Meta began publicly positioning itself as a people-focused company. In an advertising campaign, it said it was “betting on people,” while Zuckerberg presented a vision of the future of artificial intelligence in which users would have greater opportunities to create their own agents.

In an internal message in June, Zuckerberg claimed that Meta was the only major company focused on empowering people and putting the power of the new technology in the hands of billions of users, rather than primarily automating work.

We are the only major company focused on empowering people and putting the power of this new technology in the hands of billions of people across all our products, rather than primarily automating work.

– Mark Zuckerberg

At the same time, Zuckerberg’s wording about there being no further “companywide” cuts “this year” left employees room for speculation. They feared that Meta might continue reducing its workforce team by team or dismissing people based on performance reviews.

Pressure on the company is being intensified by its massive artificial intelligence spending and investors’ scrutiny. This year, Meta plans to spend at least $130 billion on chips and other AI infrastructure. Analysts estimate that these expenses could nearly exhaust the company’s operating cash resources in 2026.

In his essay on the future of artificial intelligence, Zuckerberg suggested that there could be many jobs in the future, even though individual companies might have fewer employees.

Companies may become smaller – just as happened during the transition from industrial giants to technology companies. But that does not mean there will be fewer jobs overall. It means there will be more companies, each employing fewer people.

– Mark Zuckerberg

Project Organizational Transformation showed how quickly Meta was trying to rebuild its operations around artificial intelligence. But employee resistance, weaker-than-expected results from autonomous agents, and high costs forced the company to abandon the most sweeping part of the plan. Instead of rapidly replacing people with artificial intelligence, Meta was left with a complicated and unfinished experiment in the future of its own workforce.