The survey, which polled 300 senior engineering leaders responsible for mission-critical software, primarily in C/C++ environments, highlights a stark imbalance in how engineering teams spend their time. On average, teams dedicate 9.8 hours per week to writing code, but a disproportionate 16.9 hours per week to debugging issues identified during development or reported by customers in production. This debugging workload accounts for 42% of the average working week.
A central theme of the report is a growing comprehension gap. As AI generates an increasing share of production code, engineers report a loss of inherent understanding of their codebases. The survey found that 35% of AI-generated code reaches production before the team fully comprehends how it functions or impacts existing systems. This is supported by the finding that 80% of respondents believe coding agents struggle to solve difficult problems within complex codebases.
The consequences identified in the report are stark. 81% of teams experienced a production incident or outage in the preceding six months. 93% reported an instance where an AI tool's root-cause diagnosis was incorrect due to hallucination. 91% reported test escapes or serious defects entering production.
Ultimately, 79% of engineering leaders stated that while AI agents are significantly faster at generating code, the downstream effort required to understand, debug, and maintain that code means the overall release cycle is no faster than before.
Greg Law, founder and CEO of Undo, which commissioned the survey, noted that engineers lose days trying to unravel what went wrong and why, especially with code that is almost, but not quite right. Law added that while agents are great at writing reams of code quickly, they are less capable at debugging it.