The current impact of AI on engineering velocity: What 400 companies are seeing (Abi Noda & Brian Houck)

AI adoption in software engineering yields modest productivity gains, with organizations reporting a median 8% increase in pull request throughput rather than the anticipated exponential growth. Because coding occupies only 14% of a developer's total time, AI-driven improvements to code generation have a limited impact on overall organizational velocity. Higher performance gains correlate with an "all-in" cultural shift and centralized leadership advocacy rather than tool usage alone. As engineering teams move beyond basic AI integration, they must address the "false velocity" trap—prioritizing superficial output metrics over sustainable product innovation. Emerging strategies now focus on measuring "agent experience" and optimizing the entire software development lifecycle, including tasks beyond coding, to better align AI deployment with measurable business outcomes.