AI coding tools shift developers towards review work

BairesDev has released survey data showing developers are spending less time writing code as artificial intelligence tools take on a larger share of programming work. The findings point to a shift in software jobs towards review, debugging and learning new tools.

The company's Q3 2026 Dev Barometer drew responses from 705 software developers across more than 60 countries and 41 Chief Technology Officers from Fortune 500 and mid-market organisations. It compared those responses with data gathered a year earlier to track how day-to-day work has changed as AI coding tools have spread across software teams.

The survey found that the share of developers who said AI now writes at least half their code rose to 42% from 12% a year earlier. Developers also said they save an average of 13 hours a week on coding, up from about seven hours in the earlier survey.

That reduction in manual coding did not translate into lighter workloads. Only 21% of developers said they now spend more than half their week writing new code from scratch, while 67% reported spending more time reviewing AI-generated code and 52% said they spend more time fixing problems introduced by AI systems.

Work shift

The data suggests the developer role is moving away from direct code production and towards oversight. Developers also said they now spend an average of nine hours a week learning AI tools and new technologies, compared with four hours in the Q3 2025 survey.

The survey found that 86% of respondents described their role as more fulfilling because of AI, up from 76% a year earlier. Among developers who received pay rises, AI tool fluency was the most commonly cited factor at 29%, followed by system design and architecture at 20% and human skills such as communication, mentorship and cross-functional collaboration at 15%.

The results come as large technology groups continue to introduce coding assistants and software development tools built on generative AI. Meta's launch of its AI coding tool Muse Code adds to evidence that AI systems are becoming a routine part of engineering workflows, even as questions remain about code quality, accountability and the skills software teams will need.

Accountability burden

The survey indicates that responsibility for code produced with AI still rests mainly with human developers. It found that 78% of Chief Technology Officers reported increased spending on code review, quality assurance and validation to support AI-generated work.

Only 7% of developers said the decision to ship code had been entirely delegated to AI without human input. That points to a limited appetite among companies to remove engineers from final approval, despite faster code generation and wider adoption of automated tools.

A broader surge in spending provides context for that trend. Gartner forecast in May that worldwide AI spending would reach USD $2.59 trillion in 2026, up 47% year on year. The BairesDev findings suggest that some of that spending inside software teams is now going towards checking and validating machine-generated output rather than simply adding new tools.

The survey also highlights a possible gap between what employers are funding and what workers say is rewarded. While developers who received pay rises pointed to human skills as one of the leading factors behind better compensation, only one in four Chief Technology Officers said they were actively investing in human skills development.

Budgets instead remain focused on AI tool fluency and data infrastructure, according to the survey. That could leave companies underinvested in areas such as communication, mentoring and collaboration at a time when engineers are increasingly expected to judge, explain and manage code generated by machines.

A year of quarterly tracking has also given BairesDev a longer view of the shift. Since 2025, its Dev Barometer initiative has collected responses from more than 5,200 engineers across more than 75 countries, providing a wider dataset on how software work is changing as AI systems become embedded in development teams.

For businesses, the findings suggest that productivity gains from AI coding tools may not reduce headcount or working time in the way some early expectations implied. Instead, they may redirect labour towards quality control, problem-solving and keeping pace with a fast-changing toolset.

For developers, the report paints a picture of a job that is being redefined rather than removed. More of the routine drafting of code is being handled by AI, but more responsibility is also being placed on engineers to decide what should be kept, corrected or rejected.

Chief Executive Officer Darren Shimkus said the time developers save on coding is being redirected rather than returned. "Developers nearly doubled the time AI saves them in coding, to 13 hours a week. Not one of those hours came back. Reviewing check-ins. Fixing what the model broke. Learning next quarter's tool. A year ago we read the first seven hours as capacity, and we got that wrong. The time mostly moved up the stack, into work that takes more judgment than the coding it replaced."

Shimkus also said wider AI use had increased the review burden. "In Q3 2025, 12% of developers said AI wrote half their code. Today it's 42%, and the decision to ship it still sits with one engineer, for far more code than it used to. That's why 78% of CTOs increased spending on review and validation. Every one of those dollars buys the same thing, which is an accountability layer that is currently one person deep."