Meta, the social media giant, has embarked on an ambitious organizational experiment with its new applied AI engineering team, revealing a management structure that significantly deviates from established norms. This division, tasked with spearheading the company’s superintelligence initiatives, reportedly operates with a staggering 50-to-1 employee-to-manager ratio. This figure doubles what is typically considered the upper limit for effective oversight, often cited as a 25-to-1 span-of-control. The move has prompted considerable discussion among organizational behavior experts, with some expressing strong reservations about its potential outcomes.
The concept of a flatter organizational structure, where fewer managers oversee larger teams, is not new. Proponents argue that it fosters agility, streamlines decision-making, and brings management closer to the front lines, theoretically enhancing innovation through improved cross-functional collaboration. This model is believed to boost employee engagement and a sense of ownership, as individuals feel more connected to authority figures. Indeed, Meta is not alone in exploring such structures. A recent Gallup report from January indicates a broader trend across U.S. companies, showing an increase in the average number of direct reports per manager from 10.9 in 2024 to 12.1 in 2025. This represents nearly a 50% rise in team size since Gallup began tracking this metric in 2013, with a notable two-percentage-point increase in teams comprising 25 or more employees.
However, the extreme ratio adopted by Meta’s AI team raises particular concerns. André Spicer, executive dean of Bayes Business School in London and a professor of organizational behavior, has voiced significant apprehension, suggesting the approach could “end in tragedy.” While acknowledging that flat structures can offer short-term cost savings and bolster quarterly financial reports, Spicer warns of potential medium-term problems. He notes that such models are generally most effective in “expert-oriented organizations” like software engineering and academia, where peer coordination and professional norms often guide work. Even within these suitable environments, challenges can arise.
One primary concern centers on the potential for junior or less experienced employees to be overlooked in such a expansive structure. With managers spread thin, the focused attention needed for development and guidance could diminish. Simultaneously, the sheer volume of direct reports risks overwhelming line managers, leading to burnout. A lack of clear direction for many team members is another anticipated issue, which could result in managers’ limited time being disproportionately consumed by “the loudest people or the problem cases,” effectively neglecting the majority. Spicer suggests that, in the absence of formal hierarchies, large flat teams often spontaneously develop informal leadership structures, with research pointing to an optimal team size of around seven people per manager.
The history of organizational flattening offers a cautionary tale. The computerization wave of the 1980s and 1990s similarly led to a significant reduction in middle management roles. Yet, this trend eventually reversed as companies grew more complex, necessitating the reintroduction and expansion of middle management layers to serve diverse stakeholders. Spicer highlights that, despite past delayering efforts, there has been an “explosion of middle management” from the 1980s to the present day. While advancements in AI might offer solutions by automating tasks such as allocation and employee counseling, potentially easing the burden on managers in flat structures, Meta has yet to disclose how its AI engineering team will navigate these inherent challenges. The company’s bold experiment will undoubtedly serve as a critical case study in the evolving landscape of workplace organization.
