New Survey Reveals Challenges in Building Developer Confidence in AI-Generated Code

As the world of software development evolves, a significant transformation is underway with AI taking center stage. A recent survey conducted by Sonar, a leader in agentic code verification and governance, reveals intriguing insights about the state of developer confidence in AI-generated code. The 2027 State of Code Developer Survey, which pooled responses from 1,026 developers and engineering managers globally, sheds light on the burgeoning reliance on AI and the hurdles that come with it.

The Rise of AI in Coding

The survey reports that AI now plays a pivotal role, generating or substantially assisting with an impressive average of 65% of the code that developers commit. Not only is AI becoming an integral player in the coding process, but it is also expected to become even more prevalent. A staggering 34% of respondents anticipate that their teams will either develop or utilize agentic software factories within the next year, a sharp increase from just 5% currently engaged in such practices. This growth indicates a significant shift towards autonomous coding practices, driven by the capabilities of AI.

However, this rapid adoption brings a pressing concern to light: the ability to verify AI-generated code in a timely manner. The survey uncovered that an overwhelming 76% of developers feel they create AI code faster than they can verify it. The consequence? A substantial 53% report that they now spend more time reviewing code than actually creating it, a considerable rise from the 18% doing so prior to adopting AI tools.

The Verification Gap

Despite the presence of AI tools, the scrutiny of AI-generated code remains inconsistent. A mere 39% of the respondents fully agree that AI-generated code undergoes thorough checks before being committed. Alarmingly, 94% of developers express that they do not entirely trust AI-generated code, pointing to a crucial need for stronger verification processes. Among those surveyed, 42% voiced significant concerns over the challenge posed by well-written AI code, which they believe makes bugs more challenging to detect.

Yet, even with these apprehensions, only 23% of participants reported having well-defined guidelines or automated checks specifically designed for AI-generated code. Given that 76% of developers utilize AI tools for critical services, fostering code confidence has become an indispensable requirement rather than a mere luxury.

Building Trust in AI-Driven Code

Sonar’s survey identifies four cornerstone strategies developers can adopt to enhance confidence in agentic development:

1. Intent: Clearly define project outcomes before initiating the coding process. The survey highlights that 37% of respondents attribute unclear requirements as a major source of frustration, while 26% struggle with correcting or rewriting AI-generated code. Establishing well-defined business objectives, architectural standards, quality criteria, and security requirements at the outset can provide a clear target for coding agents and assist teams in evaluating their work.

2. Guardrails: Imposing guidelines and governing the actions of coding agents as they work can significantly improve outcomes. Teams that set forth explicit standards and controls are more likely to achieve a return on investment (ROI) of 50% or more when compared to those with fewer controls. Equipping agents with relevant context and restrictions aids them in adhering to coding standards while maintaining necessary limits on tool access and permissions.

3. Verification: Implementing strong verification tools is crucial for boosting developers' confidence in AI-generated code. A remarkable 77% of respondents noted that robust verification systems increase their trust in AI outputs. Checking and adjusting each change independently before release will help to ensure that the code meets established standards, rather than accepting it solely based on superficial correctness.

4. Evidence: Measurability is essential for cultivating confidence in AI’s contribution. Currently, just 36% of developers are able to track AI's impact on productivity effectively, with only 22% having a clear understanding of the value derived from AI coding efforts. Documenting outcomes and maintaining records of what was reviewed, rectified, and verified will enable teams to assess the true value of AI involvement in coding processes and make informed release decisions.

The Path Forward

According to Tariq Shaukat, CEO of Sonar, “AI has made code abundant. The challenge now is building confidence in what ships, as evident by the responses in our latest State of Code report.” As the reliance on AI in software factories escalates, teams cannot merely depend on human oversight to scrutinize every line of code at the end of the process. Instead, engineering confidence within the development framework becomes paramount. This requires a clear intent, guiding frameworks, independent verifications, and documented evidence of code quality. The ultimate aim is not just to produce a greater volume of code but to ensure that more secure and reliable software is deployed, thereby delivering tangible business value.

For those interested in delving deeper into the findings, the complete 2027 State of Code Developer Survey report is available at Sonar's official site. This research underlines the current landscape of software development and provides valuable insights for teams navigating the challenges of modern coding practices.

About Sonar

Sonar empowers organizations to confidently deploy agentic code. With a client base that includes 75% of the Fortune 100, Sonar effectively reduces outages and enhances security while minimizing the costs and risks associated with agentic software development. Their flagship platform, SonarQube, meticulously verifies over 750 billion lines of code daily, ensuring vulnerabilities and bugs are addressed before moving to production. Trusted by over 7 million global developers, including teams from Nvidia, ServiceNow, Booking.com, Goldman Sachs, AstraZeneca, and Ford Motor Company, Sonar is positioned as a leader in the realm of AI code verification and governance. To discover more about Sonar, visit their website at sonar.com.