Research from The Adaptavist Group shows that 83% of developers are using AI-assisted vibe coding, with growing concern about governance and team coordination.
The survey of 240 professional developers in the US and UK found AI-assisted coding is now common across software teams. It also found that 87% of respondents reported time savings, 74% said they were building more than before, and 71% said the approach made them better at their job.
Those gains, however, are being matched by pressure on review processes and day-to-day management. The research found that 71% of developers said vibe coding had increased coordination work for their team, while 63% said it had made planning and task tracking more complex.
Many respondents said they were relying more heavily on project management software to cope with the extra workload. According to the findings, 73% said vibe coding had increased their appreciation for project management tools, and the same share said they were using those tools more often since adopting AI in development work.
Review gaps
The figures suggest internal controls are not keeping pace with usage. While 82% of developers said governance and structured review processes were essential, 11% said they deploy AI-generated code without human review every time, and 30% said that happened at least some of the time.
Nearly half of respondents, 47%, said there was only a basic additional review process for AI-generated code. Another 14% said no additional review process existed at all.
The survey also pointed to a visibility problem for employers trying to set standards. Two in five developers, or 40%, said they do not always fully disclose how extensively they use AI tools at work, citing concerns about how that usage may be perceived or governed.
The findings come amid a broader debate across the software industry over whether AI coding tools are improving productivity faster than organisations can adapt their oversight. In the research, vibe coding refers to a development approach in which engineers rely heavily on AI-generated code and high-level prompts, with less emphasis on detailed hands-on implementation and deep understanding of the resulting software.
Team effect
The data challenges the idea that AI coding tools mainly benefit individual developers working alone. Instead, it suggests that higher output creates more work for teams responsible for planning, alignment and review.
Neal Riley, AI Innovation Lead at The Adaptavist Group, highlighted the operational effect of that shift.
"Developers are shipping more, and shipping faster," Riley said.
"Productivity is no longer the limiting factor. When developers can produce three times the code in half the hours, and code generation accelerates, organisations need better visibility, planning, alignment, review, and orchestration. The teams that win in this era won't be the ones with the fanciest AI tools. They'll be the ones with the strongest connective tissue around them."
Developers also identified risks beyond process management. More than half, or 55%, said vibe coding introduces technical debt, while 33% said it creates operational risks such as downtime or software failures.
The research also raised concerns about skills development. Some 67% of respondents said vibe coding limits learning opportunities for junior developers, and 39% said they worried it could threaten their own job security.
Even so, sentiment towards the tools remained broadly favourable. The survey found that 74% of developers said vibe coding made their work more enjoyable, 67% said it improved creativity, and 76% supported deploying vibe coding tools across their organisation.
Next stage
The survey suggests many developers already see vibe coding as a step towards more autonomous software creation. Nearly half of respondents, 49%, said they believed vibe coding would eventually be replaced by agentic engineering, which the research described as a more autonomous and workflow-driven approach.
Another 30% said that shift could make coding even faster and more efficient than vibe coding is today. The findings suggest that while developers are largely convinced by AI's productivity gains, many organisations still face a basic management problem: how to review, govern and track a larger volume of software work without slowing teams down.
For now, the clearest message from the research is that adoption is running ahead of control. Even among developers who support AI-assisted coding, the data shows a growing expectation that employers must tighten review practices and improve coordination as output rises.
Among those surveyed, 82% said governance and structured review processes are essential.