TLDR: A 2025 report by Harness highlights the growing developer crisis, with 78% spending 30% of their time on manual tasks. While AI-powered code generation offers efficiency, it also increases workload due to debugging, security, and quality assurance. The report identifies significant concerns among engineering leaders and developers regarding vulnerabilities, performance issues, and ‘Shadow AI,’ emphasizing the need for comprehensive AI integration across the entire software delivery lifecycle to truly augment developer capabilities rather than replace them.
The software development landscape in 2025 is marked by an intensifying developer crisis, as revealed in a recent report by Harness. Market demands have fundamentally reshaped the developer’s role, moving beyond mere code writing to an expansive scope encompassing security, operations, performance, and user experience. This consolidation of responsibilities has significantly increased cognitive load, exacerbated by a prevalence of legacy processes and repetitive tasks.
The report indicates that a staggering 78% of developers dedicate at least 30% of their time to manual, repetitive tasks such as compliance policy writing, quality assurance testing, and error remediation. For organizations with 250 or more developers, this translates to an annual loss of over $8 million in productivity per engineering team, based on an average developer salary of $107,599. Beyond financial implications, this ‘developer toil’ contributes to burnout, with 50% experiencing an unhealthy work/life balance, 41% increased burnout, 48% higher stress and anxiety, 42% less time with family, and 46% being enticed to leave their organizations due to frequent overtime.
In response to these challenges, AI-powered development tools, particularly code generation, have emerged as a potential solution. The report notes that 81% of engineering leaders and 83% of developers believe AI tools have made software development more efficient. However, this adoption is not without its complexities. A significant 92% of developers acknowledge that while AI tools increase code volume, they also expand the ‘blast radius’ of problematic deployments. Contrary to initial expectations, developers are spending more time debugging AI-generated code and resolving security vulnerabilities. The adoption of AI code generation has, in many cases, increased developer workload due to new demands in code review, security validation, and quality assurance, potentially offsetting initial productivity gains.
Engineering leaders and developers share notable concerns regarding the increased use of AI. Specifically, 52% of leaders and 54% of developers foresee an increase in vulnerabilities and security incidents. Performance problems are a concern for 52% of leaders and 51% of developers, while 46% from both groups anticipate more manual downstream work. Regulatory non-compliance risk is a worry for 44% of leaders and 41% of developers, and 40% of leaders and 37% of developers expect a reduction in code quality.
A critical issue highlighted is ‘Shadow AI,’ where unauthorized AI codegen tools create significant IT challenges. This raises serious compliance and intellectual property concerns, as sensitive code snippets may be unknowingly shared with external AI services without proper governance. Alarmingly, 50% of developers do not use the AI tools provided by their IT departments. The report identifies critical gaps in internal AI tooling policies, with 50% lacking processes for assessing code vulnerabilities, 58% without specific safe/unsafe use cases, 43% missing guidelines on code input, and 42% without clear rules on approved tools.
The report emphasizes that the full benefits of AI-assisted software development, agreed upon by 95% of engineering leaders and 96% of developers, will only be realized when its application extends beyond mere code generation to the entire Software Delivery Lifecycle (SDLC). Over the next 12 months, engineering leaders plan to invest in AI across various areas: continuous integration and delivery/deployment (50%), performance optimization (48%), security and compliance (42%), ensuring cost-optimized code (40%), code generation (39%), QA/testing (28%), and error remediation (26%).
Also Read:
- Integrating Generative AI in Healthcare Software Development: Lessons from a Real-World Project
- Student Trust in AI Coding Tools: An Initial Surge Followed by Realistic Assessment
Despite concerns about AI’s impact, the report finds that 0% of respondents are concerned about AI tools replacing developers. Instead, the role of developers is expected to evolve, with AI handling routine tasks and freeing human talent to focus on higher-value activities such as prompt engineering, output validation, and capabilities integration. The integration of AI into development workflows requires careful consideration, balancing automation benefits with the crucial need for human oversight and understanding. The ultimate goal, as the report concludes, is to augment developer capabilities rather than replace them.
