AI code causes real world production outages

A survey by Undo, a debugging company, finds 81% of 300 engineering leaders linked a production incident to AI coding tools in the past six months.

Chris Stokel-Walker, a freelance journalist writing for LeadDev, reports on a survey by Undo, a company that sells debugging tools. Undo asked 300 senior engineering leaders at large US and UK companies, in July and August 2026, how AI coding tools affect their work. Most reported faster code but no faster releases.

• 81% of the leaders said their organisation had a production incident or outage linked to AI coding tools in the last six months.
• Engineers spend an average of 16.9 hours a week debugging and 9.8 hours writing code, according to the survey.
• 79% said release cycles are no faster than before AI, and 35% of AI-generated code reaches production before the team fully understands it.
• Undo's report recommends defect rates and mean time to resolution, rather than lines of code, as the measure of whether AI speeds up delivery.

An engineering team can track how long its people take to understand and debug agent-written code, alongside how much code agents produce. This is an inference and is not the author's claim. The figures come from Undo, a vendor that sells debugging software.

Derived by Working Surface from the article; more in the Playbook. Source line: AI can produce code extremely quickly, but it remains vulnerable to misunderstanding requirements and edge cases.

13 October 2026: AI agents have moved the slow step in software work from writing code to checking and running it, so teams need to measure and staff that later step deliberately.