For decades, the software has been built online by human hands. This process changes quickly due to AI according to the last annual Google Dora: Development of AI IA software Report, published today (September 23), 90% of technology professionals Now use AI in their workflows, representing a leap of 14% compared to last year. The survey of more than 5,000 software professionals and IT specialists has revealed that developers are counting on AI for tasks ranging from the writing of code extracts to the execution of tests and the security revision.
Despite a higher adoption of AI, however, confidence in technology remains low. While most say that AI makes them faster and more productive, only 24% Say that they trust him “a lot” or “a lot”. Almost a third admits that they trust him “a little” or not at all.
“Prioritization at the conference room shows that this change is probably there to stay. Each organization is faced with pressure to improve software performance, even in the face of large economic and constraint pressures, “said Nathen Harvey, principal of the study and defender of Google Cloud developers. “At the individual level, AI has captured human imagination and inspired developers to find ways to stimulate higher and lower improvements for businesses.”
The study revealed that 85% of professionals Let’s say that AI has made them more productive, although 41% call for improvement only “light”. Less than 10% reported a drop in productivity. The developers now spend a median of two hours a day using these AI tools, and the most efficient organizations report that AI stimulates flow, which allows them to provide applications more quickly and more reliable.
The quality of the code is the place where the impact of the AI is most obvious. A large part of the software he helps to create ends up working in production systems much longer than developers had never planned it. This longevity shows that the code generated by AI can be more useful than expected, but it also increases the issues. Readability and adaptability are much more important than rapid corrections when judgment of the quality of the code.
The software is based on constant code changes, such as adjustments, fixes and new features, to stay alive. The feedback loops of automated tools or users act as vital signs, signaling the health of the system. But Harvey warned that, although AI is developing development, this can also make delivery of software more unstable. “Even with the help of the AI, the teams will always need means of obtaining rapid comments on the code changes that are brought,” he said.
For the moment, the developers are hesitant to abandon control. Only a quarter in the survey indicates that they have great confidence in AI’s coding capacities. Google researchers call this the “paradox of confidence”: AI is a useful assistant, but not yet a real partner. This skepticism could slow progress towards advanced uses such as autonomous tests or fully automated versions.
Harvey noted that the developers deal with the production of AI with the same healthy skepticism that they apply to essential resources, such as coding solutions found on Stack Overflow – useful but never blindly confidence. “AI is as good as the data it has access,” he said. “If the internal data of your business is disorderly, partitioned or difficult to reach, your AI tools will give generic and unnecessary answers, holding you instead of helping.”
Harvey also noted that AI had not attenuated professional exhaustion or reduced friction. Although this stimulates individual productivity, these challenges often arise from business culture, leadership and processes – the problems that technology cannot solve alone. “If the leaders are starting to expect more because the AI makes the developers faster, it could even add to the pressure,” he added.
To fill this gap, Google introduced the Dora Ai capacities model, a framework of seven technical and cultural practices aimed at amplifying the impact of AI. The model emphasizes user development, clear communication and small lots workflows – emerging that this success requires more than new tools.
“Culture and state of mind continue to be enormous influences to help teams achieve and maintain higher performance. A climate for learning, a quick flow, rapid feedback and continuous improvement practice are what motivates lasting success. The AI amplifies the need for all these elements and provides the catalyst to transform itself as a way, ”said Harvey.
In the end, Google’s 2025 report maintains that the greatest obstacle is not adoption but confidence. Without greater confidence in the reliability of AI, the future of software development will depend as much on the faith of winning developers as on the improvement of technology itself.
