Can AI coding tools be trusted? Developers aren't so sure - over a third are concerned about AI-generated code quality despite widespread adoption and productivity gains

Over one-third (39%) of software developers have little to no trust in AI-generated code despite most admitting to using AI coding tools daily, according to new research. The study found that while over 75% of developers use AI for at least one daily responsibility, the lack of confidence in AI-generated code suggests a critical gap in effective AI integration within workflows and highlights a need for careful management of AI's role in coding.

The report revealed that while the majority of surveyed developers acknowledged some level of trust (87.9%) in AI-generated code, a significant portion still expressed concern about its impact on software delivery performance, which experienced a decrease in both throughput (1.5%) and stability (7.2%) as AI adoption rose. Developers pointed out that improving development processes does not guarantee better delivery outcomes unless foundational practices like small batch sizes and robust testing are adhered to.

In light of the findings, the report emphasized the necessity for clear guidelines governing AI use in development environments, addressing procedural concerns, and promoting open communication regarding AI's impact on workflows and project outcomes. This approach could help mitigate the downsides commonly associated with AI-generated outputs and bolster developer confidence in integrating these tools effectively.

Read at ITPro