In a revealing study titled The AI Control Gap, LaunchDarkly explores the paradoxical nature of modern software development. While artificial intelligence (AI) accelerates coding processes, it simultaneously introduces new challenges and potential hazards for developer teams.
According to the report, an impressive 94% of teams acknowledge the acceleration in code development due to AI. However, there’s a flip side: 91% of these teams have expressed apprehensions about deploying AI-driven changes to production.
The rapid delivery of code hasn't come without its pressures. A significant 81% of respondents admitted to deliberately releasing risky code in the past six months, primarily because of tight deadlines. This relentless cycle forces 38% of teams to dedicate over a quarter of their time handling incidents.
The most profound impacts on software delivery identified by the report include:
- 42% report reduced morale and increasing burnout.
- Despite 99% having runtime guardrails, approximately 70% need to roll back weekly.
- 69% engage in hotfixing or rollbacks at least once a week.
- 83% report AI-generated code as likely to cause production issues.
Additionally, the analysis notes that over 60% of firms experience customer-visible problems such as performance degradation, failed service level agreements (SLAs), or customer churn. Alarmingly, 45% revealed that slow or risky releases have cost them market opportunities.
According to Joe Byrne, Global Field CTO at LaunchDarkly, the findings encapsulate the AI Control Gap: the disparity between the rapidity with which AI enables code shipment and the safety of deploying such code in real-time productions. He emphasises, "The tension between speed and control is becoming one of the biggest challenges in modern software delivery."
With 77% of teams still grappling with the fast-paced nature and potential pitfalls of AI-accelerated development, there's a burgeoning case for prioritising resilience. Enterprises must champion strategies focusing on controlled releases and runtime stability, aligning these with core business metrics like speed and cost efficiency.
