Meta CEO Mark Zuckerberg explored a sweeping restructuring plan earlier this year that could have reduced the size of some company teams by as much as 60% as part of an effort to make the Facebook and Instagram parent an “AI-native” organization. Code-named Project OT, for Organization Transformation, the initiative envisioned AI agents taking over a large share of routine work while smaller, highly skilled groups of employees supervised the systems. Meta ultimately proceeded with a 10% workforce reduction in May but canceled a second restructuring wave that had been planned for November.
The reversal came after employee resistance intensified and internal evidence raised questions about whether AI agents were delivering the productivity improvements required to justify such a dramatic reorganization. Internal reports cited by Reuters indicated that AI-assisted coding increased output in some areas but did not produce comparable gains in overall productivity, while technical and security problems reportedly increased. Meta has confirmed Project OT existed but stressed that the 60% figure applied to scenarios for some teams, not to the company’s entire workforce, and that the plans included redeployments and unfilled positions as well as layoffs.
Project OT Was Designed To Make Meta “AI-Native”
Project OT represented a much more ambitious strategy than a conventional cost-cutting exercise. Meta’s leadership envisioned an organization in which AI agents would perform much of the routine work previously handled by human employees, allowing smaller teams to supervise those systems and concentrate on higher-value tasks.
The plan emerged after Meta executives studied AI-native organizational models and experimented with smaller teams, or “pods.” The objective was to reduce management layers, accelerate product development and allow employees supported by AI to produce more with fewer resources.
How The Proposed AI Organization Would Work
| Project OT Element | Proposed Approach |
|---|---|
| Core philosophy | AI-native organization |
| AI agents | Handle significant routine workloads |
| Human employees | Smaller, highly skilled teams |
| Team structure | Smaller “pods” |
| Management | Fewer layers |
| Product development | AI-first workflows |
| Workforce changes | Layoffs, redeployments and role closures |
| Most aggressive scenario | Up to 60% reduction in some teams |
| Restructuring waves | May and November |
| Company-wide 60% cut | Not planned, according to Meta |
The distinction between a 60% reduction in some teams and a 60% reduction across Meta is critical. Meta said it never intended to eliminate 60% of its entire workforce. The most aggressive scenarios applied only to certain teams and involved multiple forms of workforce adjustment.
Zuckerberg Ultimately Canceled The Second Layoff Wave
Project OT was designed to be implemented in two major waves.
The first was scheduled for May, while a second restructuring was planned for November. Meta ultimately carried out a 10% workforce reduction on May 20, but Zuckerberg reportedly canceled the November phase just hours before the first round of layoffs.
The sequence is significant because the company had already committed to the initial layoffs when Zuckerberg changed course on the broader restructuring.
Meta’s 2026 Workforce Restructuring
| Development | Timing / Scale |
|---|---|
| Project OT developed | Early 2026 |
| First restructuring wave | May 2026 |
| Workforce reduction | About 10% |
| Second planned wave | November 2026 |
| November restructuring | Canceled |
| Some team scenarios | Up to 60% reduction |
| Previous major Meta cuts | About 25% of workforce, three years earlier |
Meta said Project OT was also intended to move thousands of employees into newly established priority teams, including work related to generating training data for its AI models.
The company therefore did not view the initiative exclusively as a headcount-reduction exercise. It was also an attempt to redirect people toward areas considered strategically important to Meta’s AI ambitions.
Employee Backlash Put Pressure On The Plan
One of the biggest problems was internal resistance.
Employees increasingly believed that the AI transformation was being used to replace human workers rather than simply make them more productive. That perception intensified after reports emerged about layoffs and other AI-related workplace initiatives.
An internal Pulse survey cited in reporting showed employee sentiment falling from 74% favorable to 55% favorable during the period.
The company’s efforts to train AI systems also contributed to frustration. U.S. employees were reportedly asked to install software capable of tracking keyboard and mouse activity so that the data could be used to train AI agents to mimic human computer use. The initiative generated significant criticism internally.
Employee Sentiment Shift
74% Favorable
↓
Project OT + Layoff Reports + AI Tracking
↓
55% Favorable
The decline illustrates one of the central difficulties of an AI-led organizational transformation: employees may become less willing to adopt new tools if they believe the technology is primarily being developed to eliminate their jobs.
AI Productivity Gains Were Not Strong Enough
The second major problem was economic and operational.
Meta’s internal data reportedly showed that employees were using AI more heavily and that certain measures of software output had increased. But those gains did not translate into the level of productivity improvement that would have justified the scale of the restructuring.
For example, code changes to Meta’s internal software platforms and infrastructure reportedly increased 220% year over year, while changes that resulted in new or upgraded features reaching users increased only 36%.
| Productivity Indicator | Reported Change |
|---|---|
| Code changes to internal platforms/infrastructure | +220% YoY |
| Changes resulting in new/upgraded user features | +36% YoY |
| Major technical/security incidents | +40% YoY |
| Employee time spent “firefighting” incidents | +70% |
These figures highlight the difference between producing more code and producing more useful products.
AI can make it easier for engineers to generate code, but the resulting output still needs to be tested, integrated, secured and maintained. If the additional code creates more problems than value, higher output does not necessarily mean higher productivity.
Technical Problems Complicated The AI Transformation
Internal reports also indicated that AI-related systems were contributing to technical and security problems.
Major technical and security incidents reportedly increased 40% from the previous year, while employee time spent resolving those incidents rose as much as 70%. Meta did not comment on those internal figures.
This created a potentially difficult cycle.
More AI-Generated Output
→ More Systems And Changes
→ More Unexpected Problems
→ More Human “Firefighting”
→ Less Net Productivity
The issue is particularly important for a company as large as Meta, where software changes can affect billions of users and multiple interconnected platforms.
A productivity tool that works well for a small startup may create very different risks when deployed across systems supporting Facebook, Instagram, WhatsApp and Meta’s advertising infrastructure.
Some Engineering Teams Already Shrank By 30%
Even before the second Project OT wave was canceled, the combination of layoffs and employee redeployments had already reduced staffing in some engineering units.
According to internal documents and people familiar with the planning, headcount in some engineering teams had fallen by as much as 30% by the end of May.
This means the cancellation of the November restructuring did not restore the organization to its earlier structure.
Instead, Meta retained some of the smaller-team and AI-focused changes while abandoning the most aggressive additional reductions.
By June, at least 11 units, including engineering and research teams, had implemented smaller pod-based structures, according to internal announcements and people familiar with the arrangements.
Zuckerberg Did Not Abandon AI
The decision to halt the second wave should not be interpreted as a retreat from artificial intelligence.
Meta remains one of the world’s largest corporate AI investors. Reuters reported that the company expects to spend $130 billion or more on AI infrastructure and chips in 2026, while investors have continued to scrutinize the scale and returns of that investment.
Zuckerberg has instead shifted the public emphasis toward AI as a tool for empowering employees.
In internal communications and a broader public campaign, Meta has promoted the message that it is “betting on people,” emphasizing AI that helps individuals rather than primarily replacing workers.
Meta’s Strategic Shift
| Earlier Project OT Vision | More Recent Positioning |
|---|---|
| Smaller teams | Smaller teams where useful |
| AI agents replacing routine work | AI assisting employees |
| Aggressive workforce reduction | More selective restructuring |
| Two major layoff waves | First wave completed; second canceled |
| AI-native organization | AI-powered human workforce |
| Automation as primary objective | Productivity and empowerment |
The distinction may prove important as Meta continues developing autonomous AI agents.
What The Project Reveals About AI And Jobs
Project OT provides an unusually clear example of the gap between AI’s technical potential and its practical deployment inside a large corporation.
AI agents may eventually be capable of completing increasingly complex workflows with minimal human intervention. But companies still need to manage reliability, security, accountability, employee adoption and the quality of the resulting work.
For Meta, the experience suggests that simply reducing headcount in anticipation of AI productivity can create risks if the underlying technology has not yet reached the required level of reliability.
The company can still pursue long-term automation while allowing the workforce to adapt gradually.
The Bigger Picture
Meta’s Project OT shows how difficult it can be for a large technology company to move from experimenting with AI to reorganizing its entire workforce around autonomous systems. The plan’s most aggressive scenarios envisioned reductions of up to 60% in some teams, but employee resistance, weaker-than-expected productivity improvements and operational concerns contributed to the cancellation of the second restructuring wave.
The episode does not mean AI-driven workforce transformation is over. Instead, it suggests that the transition may happen more selectively than the most aggressive early scenarios anticipated. Meta continues to invest heavily in AI, but the immediate emphasis appears to be shifting toward using AI to amplify smaller groups of employees rather than replacing large portions of the workforce at once.
Looking Ahead
Meta is likely to continue experimenting with AI-native teams, autonomous agents and smaller organizational structures while closely monitoring their impact on productivity and reliability. The company can also redirect employees toward AI research, model training, infrastructure and other priority areas rather than relying exclusively on layoffs to reduce costs.
The broader lesson for the technology industry is that AI adoption is becoming an organizational challenge as much as a technical one. If AI agents eventually deliver substantial productivity gains, Meta could revisit deeper workforce reductions, but the Project OT experience indicates that the economics, reliability and human consequences must align before such a transformation can be implemented at scale.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
