Andrew Ng Identifies Product Management as New Bottleneck in AI-Accelerated Software Development

Andrew Ng, a prominent figure in artificial intelligence and co-founder of AI Fund, recently highlighted a significant shift in software development, stating that AI coding assistants have accelerated engineering to the point where "what to do" has become the new bottleneck. This "project management bottleneck," as he termed it, was a key theme in his recent discussions, including a talk at the AI Startup School in San Francisco on June 17, 2025. Ng emphasized that the core of product management lies in deciding what to build, demanding product managers with heightened user empathy and rapid decision-making capabilities.

The surge in efficiency from AI coding tools means engineers can now execute tasks at an unprecedented pace, often making traditional development cycles significantly faster. Ng drew an analogy to the invention of the typewriter, which made writing easier but led to "writer's block"—where deciding what to write became the new challenge. Similarly, AI's ability to quickly generate code means the constraint is no longer coding speed but rather the clarity and speed of product direction.

Ng further elaborated on decision-making, stressing that product managers should not solely rely on user survey data for direct decisions. Instead, he advocated for using data to "optimize understanding of users," thereby improving the quality of future decisions. This approach, centered on building a robust mental model of the user, allows for the integration of multi-source data to form a comprehensive user perspective, moving beyond isolated data points.

This shift implies a changing dynamic in team structures, with some teams even proposing a higher ratio of product managers to engineers than traditionally seen. However, Ng also noted that for scenarios requiring massive, high-frequency decisions, such as programmatic advertising, reliance should be placed on automated systems rather than human intuition. For critical team decisions, developing a strong user mental model and making swift, high-quality choices remains crucial for overcoming the newly identified project management bottleneck.