Why Only 23% of DevSecOps Teams Use AI – Surprising Findings from GitLab’s 2023 Survey

Currently, only 23% of development teams have actually implemented artificial intelligence in their software development lifecycle.

This figure comes from GitLab’s "State of AI in Software Development" report, based on a June 2023 survey of more than 10,000 DevSecOps professionals.

Although adoption is modest today, the number would rise to 90% if teams planning to use AI are included: 41% of developers say they plan to use AI within the next two years, 26% plan to use it but are unsure when, and only 9% say they will not use AI.

Among respondents who plan to use AI, at least a quarter of DevSecOps team members are already using AI tools.

The majority agree that further training is needed to adopt AI at work. GitLab notes that the most prominent concern is the lack of skills to use or interpret AI outputs, and DevSecOps professionals are eager to develop and maintain AI skills to stay ahead.

Current learning resources include books, articles and online videos (49%), training courses (49%), open‑source project practice (47%), and learning from peers or mentors (47%).

According to GitLab, 65% of respondents plan to hire new talent to manage AI in the software development lifecycle to address internal skill gaps.

Most respondents (83%) also agree that implementing AI is essential for maintaining competitiveness.

Of the 23% already using AI, 49% use it multiple times daily, 11% once a day, 22% several times a week, 7% once a week, 8% several times a month, and 1% once a month.

GitLab data shows developers spend only 25% of their time writing code, with the rest on other tasks, indicating that code generation is not the sole area where AI adds value.

Other enterprise AI use cases include predicting productivity metrics, recommending code reviewers, summarizing code changes or issue comments, automatically generating tests, and explaining how to exploit vulnerabilities.

The most popular AI use cases in practice are chatbots for asking questions in documentation (41% of respondents), automatic test documentation generation (41%), and summarizing code changes (39%). Additionally, 55% of respondents are interested in code generation and code suggestions, ranking first among developer needs.

When considering AI’s impact, many developers also worry about job security: 57% fear AI will "replace programmers' roles within the next five years".

Job replacement is not the only concern; 48% worry that AI‑generated code will lack copyright protection, and 39% fear it may introduce security vulnerabilities.

Privacy and intellectual‑property concerns are also prominent: 72% worry AI access to private data could expose sensitive information, 48% fear exposure of trade secrets, another 48% are uncertain about data storage locations and methods, and 43% are unsure how to use the data responsibly.

The majority of respondents say they need to evaluate an AI tool’s privacy features before purchasing.

GitLab concludes: "Combining human team members' experience with artificial intelligence is the best—and perhaps the only—way for organizations to fully address security and intelligence challenges. In our survey data, AI can generate code faster than human developers, but human team members must verify AI‑generated code for errors, security vulnerabilities, or copyright issues before it goes into production. As AI becomes a frontier in software development, organizations should focus on optimizing AI for efficiency while balancing it with human review to ensure integrity."

“Leveraging the experience of human team members together with AI is the best way for organizations to fully solve security and intelligence problems, perhaps the only way.”

Mojo just raised $100 M, claiming AI‑loaded performance 36 000× faster than Python

Top 6 AI Coding Assistants in 2023

Best Python Tools Checklist for Building Generative AI Applications