GitLab has released its 2025 DevSecOps research report, The Intelligent Software Development Era, revealing a widening “AI Paradox” inside software teams. While AI is speeding up coding, development groups – particularly in Australia – are becoming less efficient overall due to fragmented toolchains, heavier compliance demands, and pressure to rapidly upskill.
The Harris Poll surveyed 3,266 DevSecOps professionals worldwide, including more than 250 in Australia. All respondents worked in IT operations, security, or software development. The results show a sector racing to adopt AI but struggling to manage the operational and organisational consequences, signalling that success in the next phase of DevSecOps will depend on balancing speed, security, and evolving AI-driven capabilities.
Australian teams report losing around seven hours each week to inefficient processes, with key contributors including too many tools, limited knowledge sharing, and teams relying on different systems. Sixty-seven percent of respondents use more than five development tools, and 63% use more than five AI tools — 14% higher than the global average. Eighty-two percent believe agentic AI will only succeed if implemented within a platform engineering model, highlighting growing concern around toolchain sprawl.
The report also shows AI is reshaping roles and accelerating skills demand. A large majority (82%) expect the industry to employ more engineers as AI simplifies coding; 89% say adopting AI is essential for future career security. Most respondents want greater investment in upskilling (87%), and 84% expect their roles to change significantly within five years.AI adoption is now nearly universal, with 99% using or planning to use AI across the software lifecycle. Yet organisations remain cautious: teams trust AI to handle only 39% of daily tasks without human review. Seventy-eight percent have run into problems caused by “vibe coding”, and 88% believe human creativity and innovation cannot be fully replaced.
Compliance is emerging as a major pressure point. Seventy-nine percent say AI is making compliance more difficult, and 85% report that more compliance issues are being discovered after deployment. More than a third (37%) identify AI-driven security and compliance as the top skill for career progression, and 85% expect compliance to be embedded directly into code by 2027.
Commenting on the findings, GitLab chief product and marketing officer Manav Khurana said the survey highlights the “AI Paradox” — where faster coding collides with bottlenecks elsewhere in the software lifecycle. He warned that fragmented toolchains are slowing developers down, and AI agents are compounding the issue. Organisations, he said, need a new framework capable of matching the pace of modern software development while addressing AI orchestration, governance, and compliance challenges that individual point tools cannot solve.
You can read the full report here.
