How Microsoft sees engineering bottlenecks changing with AI

The episode begins with host Brian Houck introducing Tim Bozarth, a leader in Microsoft's CoreAI division. Tim explains that Engineering Thrive is a framework designed to understand and improve developer productivity by focusing on outcomes like speed, ease, and quality. The goal is to create environments where engineers can thrive by making it fast and easy to do great work. The framework promotes a common language and measure for engineering success, emphasizing outcomes over activities.

Tim discusses how Microsoft has adopted Engineering Thrive over the past three years, leading to significant cultural and infrastructural changes. The company now treats developer time as a precious resource and emphasizes that improvements must enhance speed, quality, and safety without trading one for the other ('do-no-harm principle'). This systemic approach has shifted investment strategies and decision-making processes across teams, fostering a culture rooted in outcome-based metrics.

The conversation highlights how AI is transforming the software development lifecycle, particularly in planning and validation stages, which have become bottlenecks due to increased code changes and confidence requirements. AI tools, like code review agents, are shifting focus from line-level code inspection to broader confidence-building activities involving signals like testing and deployment strategies. There’s a discussion on how AI accelerates certain phases but introduces new challenges in maintaining system confidence and understanding.

Tim reflects on how AI will change the role of engineers, emphasizing that the essence of engineering—system thinking, modeling, and understanding objectives—remains vital. He suggests that future engineers will focus more on designing system behaviors and less on low-level coding. Key skills include a maker’s mindset and a continuous evaluation mindset—measuring outcomes and experimenting with new tools—rather than purely activity-based metrics.

The episode addresses the exponential growth in AI capabilities alongside rising costs—referred to as the 'exponential cost curve.' Tim explains the necessity of designing agent-based platforms and abstractions that optimize for efficiency and scale, reducing token consumption and leveraging AI systems effectively. The conversation underscores that managing AI's economic impact requires rethinking platform engineering for agent-centric workflows.

Tim offers strategic advice for VP-level leaders, urging them to shift focus from activity metrics like PR velocity to outcome metrics such as idea-to-value and product quality. He emphasizes that measuring success through outcomes fosters better decision-making, increased engineering satisfaction, and stronger business results. The overarching message is to adapt leadership thinking to prioritize impact over activity.