Meta's AI-driven productivity surge fails to translate into workforce cuts

Just last week, Bill Gates published a troubling prediction about the future of work. The Microsoft founder warned that many jobs could disappear permanently as a result of AI, and even suggested that we may need to begin defining certain fields as “Human Reserved”, jobs that we, as a society, choose to keep in human hands even when machines become capable of performing them.

At almost the same time, an intriguing case study emerged from Meta, suggesting that the road to that future may be far more complex than it seems.

According to a Reuters investigation, Meta examined scenarios this year under Project OT aimed at transforming the company into an “AI-native” organization, in which smaller teams would oversee AI systems and agents performing a significant share of the work. Some scenarios reportedly considered workforce reductions of up to 60% in certain teams. In May, the company moved forward with a round of cuts affecting roughly 10% of its workforce, about 8,000 employees, but a second planned round was ultimately canceled. What makes this story interesting is not that Meta failed to implement AI. Quite the opposite. Precisely because Meta is one of the companies investing most heavily in building the infrastructure and organizational model for the AI era, its experience challenges one of the central assumptions of the past two years: if AI doubles an employee’s productivity, a company should simply be able to reduce its workforce accordingly. According to the internal data cited in the report, the number of code changes across Meta’s platforms and internal infrastructure jumped 220% in a single year. Yet the number of those changes that actually translated into new or improved features reaching users increased by just 36%. At the same time, the company reportedly experienced an increase in significant technological and security incidents, as well as a rise in the amount of time employees had to devote to addressing them.

These figures do not suggest that AI is unproductive. In fact, they demonstrate precisely the opposite: AI can generate output at an extraordinary pace. But this is where the productivity paradox of the AI era begins to emerge. Employee productivity and organizational productivity are not the same thing. When ten potential solutions can be generated in minutes instead of one, someone still has to decide which one is right. When far more code can be written, someone has to determine how it fits into the product. And when every employee can produce more, the organization itself must be capable of absorbing, prioritizing, reviewing and implementing all that additional output. AI does not necessarily eliminate the human bottleneck. It moves it. As the cost of production falls, the value of judgment rises. From a recruitment perspective, this represents a significant shift. By September 2026, across a growing number of technology roles, the ability to work with AI is no longer an exceptional advantage. It is becoming a baseline requirement. What increasingly differentiates employees is their ability to take the abundance generated by technology and turn it into business outcomes. As machines become better at execution, the value of people who can define the right problem, distinguish what matters from what does not, understand broader implications and connect different disciplines increases. This is precisely why Gates’ discussion of “Human Reserved” jobs is so relevant now. Before deciding which jobs should be Human Reserved, organizations first need to understand what remains Human Required. None of this is a reason for complacency. AI will make certain tasks redundant, reshape professions and, in some areas, allow companies to operate with fewer employees. Gates himself expects significant disruption to the labor market. But Meta’s experience is a reminder that this transition is not linear. A 100% increase in an employee’s ability to produce does not automatically translate into the ability to cut a team in half. And that is exactly the challenge facing managers as they look toward 2027. Until now, they have primarily asked how AI can make each employee more productive. The next question is how to build an organization capable of handling all that productivity. The companies that win in the AI era may not be those that use it to replace the greatest number of employees. They may be the ones that first recognize that the scarce resource has changed. In a world where the ability to produce is becoming almost unlimited, what becomes increasingly rare, and increasingly valuable, is the human ability to decide what is worth producing in the first place. Rinat Buchholz is the Founder and CEO of Global Teams.