Meta Reorganizes to Accelerate AI
Meta is launching a new applied AI unit with a flat structure to speed up model development and tooling. The move is designed to improve performance and signals a premium on product managers who can bridge engineering and user research in fast-paced AI environments.
The new applied AI group will be led by Maher Saba, a vice president from Meta's Reality Labs division, and will report directly to Chief Technology Officer Andrew Bosworth. This unit is designed to accelerate the translation of foundational AI research into tangible consumer products across Meta's ecosystem. The team will partner with the company's Superintelligence Lab, which was established in the summer of 2025 to focus on foundational and next-generation AI models. A key feature of the new organization is its "flat structure," with a manager-to-employee ratio of up to 1:50. This design aims to minimize management layers, reduce bureaucracy, and empower individual engineers to iterate and make decisions more quickly. The structure includes two main teams: one building interfaces and internal tools, and another focused on generating data and evaluations to improve the AI models. This reorganization is the latest in a series of structural changes within Meta's AI divisions over the past year, reflecting an intense focus on accelerating AI development. The move signals a tighter integration between research labs and product teams to shorten development pipelines and create faster feedback loops. It follows a massive AI recruitment drive, reportedly involving CEO Mark Zuckerberg personally, with some compensation packages reaching nine figures. The emphasis on applied AI aligns with Zuckerberg's 2026 roadmap, which he described as a turning point for accelerated AI development. The company's capital expenditure is projected to be between $115 billion and $135 billion in 2026, a significant increase from $72.22 billion in 2025, largely to support AI infrastructure. This shift is also reshaping the role of product managers, with some at Meta now identifying as "AI builders." Armed with AI coding tools, these PMs are moving beyond coordination to actively building prototypes, reflecting a broader industry trend of blurring roles between product management, design, and engineering. This "building-first" culture is seen as a way to turn ideas into working apps more rapidly. The new unit will be crucial for integrating AI into hardware, particularly for augmented and virtual reality products within Reality Labs. This involves optimizing AI models to run efficiently on devices like smart glasses, focusing on real-time responsiveness and battery life. The goal is to enable advanced features like real-time language translation and intelligent avatars. To fuel its models, Meta has been securing more training data, including a recent multi-year deal with News Corp worth up to $50 million annually to license content from its publications. This follows a broader strategy of developing custom silicon to handle AI workloads for tasks like ranking and recommendation systems. The reorganization also comes amid leadership changes, with some reports suggesting a move to build a parallel engineering powerhouse that bypasses the company's Superintelligence Labs, which was put under the leadership of Alexandr Wang in mid-2025. The new applied AI unit under Saba is structured to move faster and feed the core AI models with data and tooling more directly.
