The AI Coding Paradox
89% of engineering orgs have had an AI-related production incident.
Many organizations have accelerated code generation faster than they have built the systems needed to validate that output. This report, based on the results of a survey of 500 U.S. IT engineers and engineering leaders uncovers the growing gap between AI coding velocity and the systems organizations have built to validate the output.
What you’ll learn
- How often AI-generated code is causing production incidents across organizations of varying sizes, and which incident categories are most common
- Why developer confidence in AI code and developer scrutiny of AI code are both rising at the same time, and what that signals about how teams have adapted
- Where the review burden is landing, and ow AI is reshaping the review burden, and why time savings are unevenly distributed across the engineering population
- How automated gate adoption correlates with outage rates, and why the largest enterprises are the most exposed
- What reviewers are actually scrutinizing in AI-generated code, and how those concerns map to the incidents organizations are reporting in production
The data at a glance:
89% of organizations have had at least one AI-related production incident.
40% of the largest enterprises (10,001+ employees) have had a production outage caused by AI-generated code. The highest outage rate of any size bracket in the survey.
41% of developers spend more time on manual review than they did before AI coding tools existed. Productivity gains are real for many, but not for everyone.
95% of developers review AI-generated code with more scrutiny
