A global survey of 2,501 IT and DevOps professionals at organizations with more than 150 employees published today finds more than two-thirds (68%) work for organizations that have implemented artificial intelligence (AI) across some or all their software delivery workflows.
Conducted by Tricentis, a provider of a platform for testing software, the survey identifies enhanced quality and risk detection (37%), enhanced accuracy and consistency (36%) and improved test automation coverage (32%) as the top benefits of integrating AI into those workflows
Overall, 53% manage between six and ten AI or automation tools across their software development lifecycle (SDLC).
However, the survey also finds that 60% admit their application developers also regularly ship untested code into production environments.
David Colwell, vice president of AI and machine learning at Tricentis, said that as more AI-generated code is created, the volume of code that has not been tested is increasing, which in the long term is going to have a negative impact on the overall quality of the applications being deployed.
That issue may become further exacerbated with the rise of AI agents. Only 35% said they are fully prepared to govern AI agents, with security concerns (27%) and regulatory and compliance requirements (22%) being the two biggest blockers to adoption.
In general, the top business impacts of the poor quality software cited by survey respondents are security breaches and compliance failures (30%), technical debt and rework costs (28%) and loss of customer or partner trust (25%).
The main operation challenges when it comes to implementing continuous best practices for software quality are tool sprawl/operational complexity (33%), code volume scaling faster than teams can manage (28%) and gaps in AI-related skills and expertise (27%).
A full 20% of respondents also noted their organization reported annual losses of up to $5 million that were attributed to poor software quality.
Not surprisingly, the top two reasons code is not tested are pressure from senior leadership to accelerate software delivery (32%), followed closely by there being too much code to test (30%).
Unfortunately, too many application developers have too much faith in the quality of the code being generated by AI coding tools as there continues to be more emphasis on speed than quality, said Colwell. As the volume of code being generated is only going to increase in the age of AI, it’s only a matter of time before more time and attention is placed on the quality of that code, he added. Eventually, the cost of poor quality software will become too big to bear as more issues arise in production environments, noted Colwell.
Ultimately, each DevOps team will need to strike the right balance between speed and quality. The one thing to remember in the age of AI is that the business benefits of meeting software delivery deadlines are, as always, negated when the quality of the software delivered winds up having a negative impact on the business that far exceeds any perceived cost created by an application that wasn’t delivered exactly on time.