AI Speeds Up Code But Breaks Production Reliability
Short summary
New Relic surveyed 2,575 engineering leaders and found that AI has become the top driver of observability demand—but it's creating a dangerous gap. Teams are shipping AI-generated code faster than they can monitor it, leading to more production incidents even as code reviews look good on paper.
- One in four organizations deploy AI agents in production with zero monitoring, creating unmitigated operational risk when autonomous systems alter code, configs, or infrastructure without visibility.
- 62% of teams ship AI-generated code without manual line-by-line verification; 78% report more production incidents and 86% see senior engineers spending extra time fixing code, despite 94% rating the code quality high during review.
- Tool sprawl is reversing years of consolidation—organizations now use multiple AI coding tools (GitHub Copilot at 49.7%, Gemini at 40.7%, Codex at 36.6%) that generate code in incompatible formats, making source-code review unreliable and runtime observability the only trustworthy signal.
Full summary
AI has become the primary reason organizations are investing in observability tools, overtaking traditional concerns like cost management and uptime. The survey of 2,575 IT leaders found that 80% plan to deploy or are already using AI application observability within 12 months. The core problem is stark: while AI helps teams ship code faster and resolve incidents quicker, it also generates code that humans can't fully review and deploys autonomous agents that operate without visibility into their actions. This creates what the report calls "phantom velocity"—the illusion of speed that masks downstream chaos. The numbers bear this out: 78% of organizations report more production incidents after deploying AI-generated code, and 86% see senior engineers spending more time firefighting, despite 94% rating the AI-written code as higher quality during initial review.
The most alarming finding is the agentic AI visibility gap. While 93% of organizations are running AI agents in production, 24.6% do so with zero monitoring—letting autonomous systems alter code, configurations, and infrastructure without any observability. This is a ticking time bomb. The report identifies five categories of AI observability now demanding separate strategies: AI application monitoring, agentic AI system monitoring, predictive analytics, ML model monitoring, and AIOps capabilities. Meanwhile, tool sprawl is returning after years of consolidation efforts. Organizations deployed an average of 4.4 observability tools in 2025 but are now adding 5 in 2026, driven by the need for AI-specific monitoring capabilities that legacy platforms don't cover.
AI is already changing how incident response works. Organizations that wanted AI-powered troubleshooting tools last year are now running them in production—adoption jumped from 24% to 35% for conversational assistants, with 44.9% actively using AI troubleshooting. The payoff comes during investigation, where AI delivers the most value (35.8% of organizations cite this as AI's highest operational benefit) by automating root-cause analysis that used to consume hours of manual detective work. The underlying tension remains unresolved: outage costs per incident are dropping slightly (down to $73.5M from $76M), but overall outage frequency stays high, and the gap between how fast code ships and how safely it runs keeps widening.