Need smarter insights in your inbox? Sign up for our weekly newsletter to get only the things that matter to enterprise AI, data and security leaders. Subscribe now
More developers than ever are using AI tools to help and generate code.
Adoption of Enterprise AI will accelerate, but new data from Stack Overflow’s 2025 developer survey reveals important blind spots. With the installation of technical debts created by AI tools that generate “nearly correct” solutions, it could undermine the productivity gains they promise to provide.
Stack Overflow’s annual developer survey is one of the biggest reports in a given year. In 2024, the report discovered that developers weren’t worried about AI still going to work. A bit ironic, stack overflow was initially negatively affected by the growth of Gen AI, resulting in a decline in traffic in 2023, resulting in layoffs.
A 2025 survey of over 49,000 developers in 177 countries reveals a troubling paradox in corporate AI adoption. The amount of AI usage continues to rise. 84% of developers have used AI tools or used AI tools since 76% in 2024.
The AI Impact Series returns to San Francisco – August 5th
The next phase of AI is here – Are you ready? Join Block, GSK and SAP leaders to see exclusively how autonomous agents are reshaping their enterprise workflows, from real-time decision-making to end-to-end automation.
Secure your spot now – Space is limited: https://bit.ly/3guplf
“One of the most surprising findings was a significant change in AI developer preferences compared to previous years, but most developers use AI, but this year they don’t like it much and are less reliable.” “This response is surprising because Tech News’s investment and focus on AI, so we hope that trust will increase as technology improves.”
The numbers tell the story. Only 33% of developers who trust AI accuracy in 2025 have fallen from 43% in 2024 to 42% in 2023. AI favors have declined from 77% in 2023 to 72% in 2024.
However, the survey data reveals more urgent concerns for technical decision makers. Developers cite “almost correct, but not perfect AI solutions.” Meanwhile, 45% say that code generated by debug AI takes longer than expected. AI tools promise productivity gains, but in reality they could create new categories of technical debt.
The “nearly correct” phenomenon confuses developer workflows
AI tools don’t just generate obviously broken code. They generate plausible solutions that require critical developer intervention to become a production-ready. This creates particularly insidious productivity issues.
“While AI tools seem to have a universal promise to save time and increase productivity, developers spend their time dealing with the unintended breakdown of workflows caused by AI,” Yepis explained. “Most developers say AI tools don’t address complexity. Only 29% think they can handle this year’s complexity from 35% last year.”
Unlike obviously broken code that developers quickly identify and discard, “almost correct” solutions require careful analysis. Developers need to understand what is going wrong and how to fix it. In many cases, they report that writing code from scratch is faster than debugging and modifying AI-generated solutions.
Workflow disruption extends beyond individual coding tasks. The survey found that 54% of developers use six or more tools to complete their work. This adds context switching overhead to already complex development processes.
Enterprise Governance Frameworks is behind hiring
The adoption of rapid AI has improved the enterprise governance capabilities. Organizations are currently facing risks that do not fully address potential security and technical debt risks.
“Vibe coding requires a higher level of trust in the output of AI, sacrificeing code trust and potential security concerns for faster turnarounds,” Ben Matthews, senior director of engineering at Stack Overflow, told VentureBeat.
Developers largely refuse to code the atmosphere for professional work, with 77% noting that it is not part of the professional development process. However, the study reveals the gaps in how companies manage the quality of AI-generated code.
Matthews warns that AI coding tools powered by LLMS can create mistakes. He pointed out that while knowledgeable developers can identify and test vulnerable code itself, LLMS sometimes cannot even register mistakes they may produce.
Security risks exacerbate these quality issues. Research data shows that when developers still rely on humans for coding help, 61.7% “cite ethical or security concerns about code as the main reason. This suggests that AI tools will introduce integration challenges regarding data access, performance and security that organizations still learn to manage.
Developers still use stack overflow and other human expertise sources
Despite declining trust, developers have not abandoned AI tools. They are developing more sophisticated strategies to integrate them into their workflows. In this study, 69% of developers spent the past year learning new coding techniques and programming languages. Of these, 44% used AI-enabled tools for learning, up from 37% in 2024.
With the rise of vibe coding and AI, research data shows that developers maintain strong connections between human expertise and community resources. Stack Overflow remains the top community platform with 84% usage. Github continues at 67% and YouTube continues at 61%. Most importantly, 89% of developers access stack overflow multiple times a month. Of these, 35% rely on the platform, especially after encountering AI response issues.
“We’ve seen a decline in traffic, but it’s never as dramatic as some people show,” Jodie Bailey, Head of Product & Technology, told VentureBeat.
That said, Bailey acknowledged that users’ daily needs were not the same as they were 16 years ago when stack overflow began. He noted that there is no single site or company that doesn’t see a shift in where users come from and how they are currently involved in Gen AI tools. With this shift, Stack Overflow critically reevaluates the way in which we evaluate success in the modern digital age.
“The future vitality of the Internet and the broader high-tech ecosystem will only be defined by the indicators of success outlined in the 90s or early 00s,” Bailey said. “Instead, there is an increasing focus on the very important role of the professional community and individuals who meticulously share and curate data caliber, information reliability, and knowledge.”
Strategic recommendations for technical decision makers
Stack Overflow data suggests several important considerations for enterprise teams evaluating AI development tools.
Invest in debug and code review capabilities: Organizations need a code review process, as 45% of developers report increased debugging time for AI code. You need a debugging tool specifically designed for AI-generated solutions.
Maintain a human expertise pipeline: Continuous reliance on community platforms and human consultations have shown that AI tools amplify, rather than replace, the need for experienced developers. These experts can identify and fix any issues with AI generated code.
Implementing step-by-step AI adoption: Successful AI adoption requires careful integration with existing tools and processes, rather than wholesale exchange of development workflows. This allows developers to take advantage of AI strength while reducing the “nearly correct” solution risk.
Focus on AI Tool Literacy: Developers using AI tools show a favor of 88% compared to 64% of weekly users. This suggests that appropriate training and integration strategies have a significant impact on the outcome.
For companies looking to lead AI-driven development, this data shows that competitive advantages come from developing superior capabilities in AI human workflow integration and AI-generated code quality control, rather than from the speed of AI adoption.
Organizations that solve “nearly correct” problems and turn AI tools into reliable productivity multipliers rather than technical debt sources, gain significant benefits in development speed and code quality.
Source link
