42 percent of developers are now letting AI write at least half of the code they write. It’s a staggering rise from just 12 percent a year ago, notes BairesDev. The hours saved on coding are now spent reviewing the AI-generated code, leaving lingering doubts over any actual efficiency gains.
BairesDev found the results in its Q3 2026 Dev Barometer, in which the nearshore developer polled 705 developers in over 60 countries, including 41 enterprise CTOs. Respondents do point to 13 hours of saved time on coding, double the seven hours reported in Q3 2025.
However, that time is consumed rewriting and reviewing AI-generated code. Only 21 percent of developers still spend more than half of their workweek writing new code entirely on their own. 67 percent note reviewing more AI-generated code than a year ago, while 52 percent spend more time debugging problems AI has introduced. Learning new tools also requires a lot of time, as nine hours a week are spent on it instead of the four reported in the same quarter last year.
“Developers nearly doubled the time AI saves them in coding, to 13 hours a week. Not one of those hours came back,” as BairesDev CEO Darren Shimkus succinctly notes.
Review becomes the bottleneck
Among the CTOs surveyed specifically, 78 percent note spending more money on code review, QA and validation. Only 7 percent of developers are letting AI cover shipping decisions entirely.
There’s clearly a gap between use and trust, as we have noted before. Research also suggests the productivity gain many assume will exists vanish as soon as AI plans meet reality. One study found that experienced developers actually spent 19 percent more time on tasks when using AI compared to not adopting it, and another survey measured 17 percent lower skill test scores among AI-assisted developers. This was particularly true for debugging.
Security over speed
Bairesdev’s Shimkus concludes that developers should target just one AI tool and seek to master it. Right now, many are sampling multiple tools, failing to spend quality time and gain true experience on any of them. This while it can benefit them financially. Among those who received raises in 2026, 29 percent cited AI tool fluency being important for it, whereas 20 percent noted system design and 15 percent human skills. Only one in four CTOs said they invest in developing those human skills. Again, there’s clearly a gap between proper planned rollouts and disaggregated adoption.