Meta Platforms halted plans for a second wave of “AI-native” restructuring on May 19, a day before the company proceeded with first-wave cuts affecting about 10% of its workforce, according to internal planning materials and employee communications described in the supplied reporting. The reversal came after Meta had spent roughly a year modeling whether AI agents could take over enough routine work to support much smaller teams across parts of the company.
The initiative, known internally as Project OT, explored scenarios in which some teams would operate with as little as 40% of their previous staffing. Meta said the work was scenario planning rather than a fixed commitment to companywide layoffs, and that several major business units were outside the project’s scope.
By July, chief executive Mark Zuckerberg told employees that the expected acceleration in AI-agent capabilities had not materialized over the previous four months. In an internal town hall, Zuckerberg said the organizational bet had not yet paid off and that the company hoped to have clearer results within three to six months.
That admission places Meta’s workforce changes alongside a more difficult question facing large technology companies: whether increasingly expensive AI infrastructure can deliver operational gains quickly enough to justify both the capital spending and the disruption of reorganizing thousands of employees.
A plan to replace large teams with small AI-assisted squads
Project OT’s operating model was set out during a January leadership meeting in Hawaii, where internal plans envisioned AI agents handling daily tasks previously performed by thousands of workers. Human employees would oversee the systems’ output rather than complete much of the underlying work themselves.
The most aggressive scenarios would have cut staffing in multiple teams by as much as 60%, a scale comparable to or greater than Meta’s previous workforce reduction of about 25% roughly three years earlier.
Under the proposal, traditional product groups of 10 to 20 people would be replaced with “AI-native” squads of three to five workers. A conventional group could include seven to 14 engineers, along with product managers, designers, data scientists, researchers and data engineers. The smaller squads would rely on agent-assisted tools for tasks such as analysis and routine prioritization.
Internal documents also described removing layers of mid-level management. Squads would report directly to a unit lead, while performance and promotion decisions would be handled through a “village method,” according to the materials. Unit heads overseeing 30 to 50 employees would lead those decisions with support from human resources and an unnamed AI system. Meta said people, rather than the AI system, made final decisions on ratings and promotions.
At least 11 units, including engineering and research groups, had shifted toward the small-squad model by June, according to the internal materials.
Productivity figures exposed a gap in the rollout
The early results showed that AI-assisted development was increasing code activity faster than it was improving user-facing products. Andrew Bosworth, Meta’s chief technology officer, wrote in an internal post in early June that changes to internal software platforms and infrastructure had risen 220% year over year.
Over the same period, changes that reached users as new or improved features rose 36%, according to Bosworth’s post.
The figures do not establish that AI tools caused the imbalance, but they point to a familiar risk in software organizations deploying generative systems at scale. Producing code is only one stage of development. Code must also be tested, integrated with existing systems, reviewed for security issues and converted into products that work reliably for users.
Internal posts cited in the reporting warned that AI-generated code was creating reliability concerns, including references to large-scale actions that would have been unlikely to occur through manual work. Major technical and security incidents, including outages and potential data-exposure events, increased 40% year over year during the rollout, while time spent on emergency-response work rose 70%, according to the internal metrics.
In early June, attackers used an AI customer-service bot to access a group of high-profile Instagram accounts, including a disabled White House account associated with former president Barack Obama, according to the reporting. Meta declined to comment on the internal failure metrics.
The combination of higher code volume and more emergency work suggests the company was confronting a bottleneck beyond software generation: maintaining control over a rapidly changing production environment.
Capital spending raises pressure for measurable results
Meta’s reorganization unfolded as the company committed extraordinary sums to AI chips, data centers and related infrastructure. The company’s 2026 capital-expenditure guidance stood at $130 billion to $145 billion, largely tied to AI infrastructure, according to the supplied materials.
Free cash flow fell 91% year over year in the second quarter, placing greater attention on whether the spending can translate into revenue growth, lower operating costs or durable product improvements. High capital expenditure alone does not indicate financial distress for a company of Meta’s scale, but it reduces flexibility when returns from new systems take longer than management expects.
The May job actions also carried a workforce cost. Earlier reporting cited a global reduction of about 10%, or roughly 8,000 roles, alongside about 7,000 transfers to teams connected with AI workflows. Meta employed about 75,500 people at the end of the second quarter.
Following Zuckerberg’s May 19 decision, the company allowed first-wave cuts to proceed the next day while abandoning work on the planned November phase. Later internal communications described allowing some transferred employees to return to their previous teams and increasing budgets for travel, team events and snacks.
Employee resistance accompanies the organizational shift
Employee sentiment weakened alongside the restructuring. Meta’s semiannual Pulse survey showed favorability falling to 55% from 74%, according to the internal results cited in the reporting.
More than 1,600 employees also signed a petition opposing keystroke and mouse-tracking tools deployed on U.S. devices. The tools were intended to collect interaction data for AI agents before Meta later paused the tracking program.
The internal reaction illustrates a practical constraint on attempts to turn AI systems into a substitute for established teams. These programs depend on employees supplying training data, supervising outputs, repairing failures and adapting workflows. Lower trust can make that transition harder, particularly when staff believe the same systems are being used to measure their work or reduce headcount.
Meta’s decision to stop planning for the second wave does not end its AI-focused restructuring. It does show that the company’s early assumptions about agent-driven productivity are being tested against engineering reliability, organizational strain and the high cost of building the infrastructure required to run those systems.
Curious how AI can enhance trading without massive restructuring? Explore smarter automation with AI copy trading today.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
