A survey of 504 DevOps professionals published today finds that 20% are now using artificial intelligence (AI) across all phases of the software development lifecycle (SDLC).
Conducted by Techstrong Research, a sister business unit of DevOps.com, on behalf of Tricentis, the survey also finds nearly a quarter of respondents (24%) are using AI in at least one phase of application development. That narrow gap suggests that after AI is first adopted it tends to spread rapidly through the SDLC.
Additionally, the survey finds that 46% of respondents work for organizations that plan to adopt AI tools to augment DevOps teams in the next 12 months. Only 20% of respondents said their organization has no plan.
Currently, writing code (45%), debugging and defect prevention (44% and testing (40%) are the three areas where AI has been most widely employed. A full 60% of those respondents who have adopted AI report seeing increases in efficiency and productivity across their application development teams. Other benefits include reduced skills gaps (53%), reduced costs (47%) and improved software quality (42%).
Mitch Ashley, chief technology officer for Techstrong Group and a chief technology advisor for The Futurum Group, said those benefits will soon be more widely experienced as AI capabilities are added to just about every DevOps tool and platform.
AI could be the biggest change to how we create software since the initial adoption of best DevOps practices and will become an even greater driver than it is today once more models are specifically trained to reduce technical debt and improve application security, he added.
Specific areas where AI is cited for providing value include testing (60%), coding (58%), security (55%), observability (53%), software builds (53%), site reliability engineering (50%) and operations (50%).
In terms of where AI is most helpful when applied to testing, 52% cited the importance of script generation, while benefits of requirements coverage through test case selection and the ability to perform a risk analysis of code changes tied at 42% each.
Mav Turner, chief product and strategy officer for Tricentis, said testing is a critical element of any AI-augmented DevOps practices. In the case of testing, for example, AI helps detect, auto-heal and predict defects during development, as well as identifies which tests need to be run based on high risk, he said.
When coupled with low-code/no-code technology, this means that, regardless of a team’s technical expertise, AI can significantly contribute to overall software quality, added Turner.
Nevertheless, the survey finds that 86% of respondents using AI report that some or significant human verification of outputs is still required, compared to 9% reporting they have complete trust in AI output.
While AI will become pervasively embedded across the SDLC, it’s not likely DevOps engineers will be replaced. However, how DevOps teams are organized is likely to change as more generative AI agents trained to perform specific tasks become available. In effect, DevOps teams will be made up of a mix of software engineers and AI agents. Collectively, that combination should enable organizations to improve both the quality of applications and the rate at which software is being built and deployed.
The one less clear thing is to what degree AI tools might be used by individual developers and software engineers versus ones that are officially sanctioned by the organization.
Regardless of how AI finds its way into a DevOps workflow, software engineers should identify today the rote tasks within a DevOps workflow that will soon be automated using AI. The challenge and the opportunity now is to reduce the level of toil DevOps teams encounter today in a way that ultimately reduces the level of burnout those teams experience tomorrow.