Key takeaways from DX Annual, a conference for senior engineering leaders, covering AI integration across the software development lifecycle. Major themes include: code writing is only 14% of engineer time so AI gains must extend beyond coding; AI token spend is becoming a managed cost like cloud bills (one org spending $128K/week); structured training matters more than tool choice (Indeed saw 36% coding time reduction with training vs. zero without); outcome-focused metrics like feature velocity are replacing activity metrics; daily heavy AI users (4+ hours) more than double output; and non-engineers like PMs and designers are increasingly shipping code with AI tools. Data points from Airbnb, Uber, Intercom, Vanguard, Microsoft, and others are shared.
Table of contents
The bottleneck of software delivery is no longer writing codeAI token spend is top of mind and organizations are measuring it differentlyStructured training and skill quality matter more than tool choiceLeaders are incorporating more outcome-focused metrics to measure AI impactProductivity gains require daily, sustained AI usagePMs and designers are shipping code, and organizations are adaptingSort: