AI speeds software development, but governance and code review lag adoption: Study

Artificial intelligence is becoming a core part of software development, with 84% of organizations now using AI during the build phase, but governance, security and code review practices are struggling to keep pace, according to a new study by Info-Tech Research Group.

Based on a survey of 578 applications, engineering and product leaders, the report found that AI is being widely used for software analysis, design, coding and testing. However, 67% of developers said AI-generated code requires more testing than code written by humans, underscoring the growing need for stronger review and quality assurance processes.

"AI is no longer sitting at the edge of software development. It is becoming part of how teams build and deliver work every day," said Brian Jackson, principal research director at Info-Tech Research Group.

"The challenge for leaders is that faster code creation does not reduce the need for disciplined delivery. It increases the need for clear review standards, security guardrails and stronger measurement so organizations can turn AI adoption into sustainable value rather than unmanaged risk."

Despite the governance challenges, organizations are already seeing measurable productivity gains. The study found that 94% of respondents reported meaningful productivity improvements from AI-assisted development, while 83% said AI had helped reduce software defects. More than half of those respondents reported defects had been cut by at least half.

However, the research found that AI adoption is advancing much faster than organizational maturity. Only 37.4% of developers using AI during software development said their organizations had formal AI processes or governance frameworks in place.

Security also remains the biggest hurdle to broader AI adoption. Nearly 48% of respondents cited security and intellectual property concerns as their top challenge, followed by concerns over output quality (42%), skills gaps (34%), data quality (26%) and poor integration with existing tools (24%).

The report also identified legacy software as the biggest technical obstacle to AI-assisted development, with 51% of respondents saying AI tools struggle to work effectively with complex or legacy codebases. Another 46% said AI-generated code often fails quality checks before production.

Info-Tech said AI is also changing developer roles. While junior developers are benefiting most from faster coding, experienced engineers are increasingly spending their time reviewing, validating and securing AI-generated code before it reaches production.

The report also found differing perceptions across software teams. Product managers were more likely to believe AI-assisted or "vibe coding" could deliver faster releases with fewer errors, while application development teams were considerably more skeptical, reflecting the realities of maintaining code quality and production readiness.

Info-Tech advised organizations to focus on building governance around AI-assisted development by defining production-readiness standards for AI-generated code, strengthening security reviews and tracking quality metrics as AI adoption expands across the software development lifecycle.