AI now writes at least half the code for 42 per cent of developers, but most of it still needs tidying up.
BairesDev’s Q3 2026 Dev Barometer surveyed 705 software developers across more than 60 countries. It separately questioned 41 enterprise CTOs from Fortune 500 and mid-market organisations.
Only 12 per cent of developers said AI produced at least half their code in the third quarter of 2025. That has climbed to 42 per cent, while 72 per cent now use AI for at least a quarter of their code.
Developers reckon AI saves them an average of 13 hours of coding each week, nearly twice the seven hours reported a year earlier. The median saving is 10 hours, suggesting a few enthusiastic prompt botherers are pulling the average upwards.
Those saved hours have not produced a generation of developers enjoying long lunches. Only 21 per cent now spend more than half their working week writing new code from scratch.
Meanwhile, 67 per cent spend more time reviewing AI-generated code than a year ago and 52 per cent spend more time debugging problems introduced by AI. Developers now spend nine hours a week learning AI tools and other technologies, up from four hours.
BairesDev chief executive Darren Shimkus said: “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.”
Responsibility has followed the code rather than vanished into the machine. Some 58 per cent of developers report greater accountability for code they did not entirely write, while 59 per cent say spotting subtle AI-generated bugs requires greater mental effort.
Human eyeballs remain stubbornly difficult to automate. Some 41 per cent said virtually no AI-generated code reaches production without specific human review, while only seven per cent said deployment decisions had been handed entirely to AI without developer involvement.
Among the 41 CTOs surveyed, 78 per cent increased spending on code review, quality assurance or validation during the past year. AI tool fluency attracted investment from nearly eight in 10 CTOs, followed by data engineering at 66 per cent, AI quality assurance at 51 per cent and AI security at 49 per cent.
The shift is turning up in pay packets too. Among developers receiving AI-related pay increases, 29 per cent credited AI tool fluency and prompting, 20 per cent cited system design and architecture and 15 per cent pointed to communication, mentoring and other human skills.
Despite spending more time checking what their silicon colleague has broken, 86 per cent said AI had made their job more fulfilling, up from 76 per cent a year ago. Some 21 per cent use AI tools their employer has not approved, while 46 per cent said token costs or usage limits had changed how they work.
