Professional Software Developers Don't Vibe, They Control: AI Agent Use for Coding in 2025

Abstract: The rise of AI agents is transforming how software can be built. The promise of agents is that developers might write code quicker, delegate multiple tasks to different agents, and even write a full piece of software purely out of natural language. In reality, what roles agents play in professional software development remains in question. This paper investigates how experienced developers use agents in building software, including their motivations, strategies, task suitability, and sentiments. Through field observations (N=13) and qualitative surveys (N=99), we find that while experienced developers value agents as a productivity boost, they retain their agency in software design and implementation out of insistence on fundamental software quality attributes, employing strategies for controlling agent behavior leveraging their expertise. In addition, experienced developers feel overall positive about incorporating agents into software development given their confidence in complementing the agents' limitations. Our results shed light on the value of software development best practices in effective use of agents, suggest the kinds of tasks for which agents may be suitable, and point towards future opportunities for better agentic interfaces and agentic use guidelines.

• The paper explores controlled AI agent use among developers and identifies motivations such as increased productivity and quality retention.
• It employs field observations and surveys to detail strategies like contextual prompting, version control, and expert oversight.
• Results indicate that while AI agents excel in automating routine tasks, they are unsuitable for complex design decisions.

AI Agent Use for Software Development in 2025

The paper "Professional Software Developers Don't Vibe, They Control: AI Agent Use for Coding in 2025" (2512.14012) seeks to investigate the roles AI agents play in professional software development, examining motivations, strategies, task suitability, and developer sentiments. It presents findings from field observations and surveys, highlighting experienced developers' controlled usage of AI agents for enhancing productivity while maintaining software quality.

Motivations for Using AI Agents

Experienced developers value AI agents for improving development speed but not at the expense of software quality. The study identifies personal productivity and the retention of software quality attributes as critical factors motivating the incorporation of AI agents in software processes. Developers appreciate agents' potential to simplify mundane tasks but remain cautious about delegating responsibilities that could impact software quality or design integrity.

Strategies for Agent Use

Developers employ strategies to oversee software design and implementation actively. Observations reveal that developers leverage agents for task execution while exercising control through meticulous review and verification of agent-generated code. Strategies include:

• Prompting Techniques: Developers use clear, context-rich prompts to guide agent actions effectively. Techniques such as step-by-step instructions, contextual file references, and proactive error handling are noted.
• Control Through Expertise: Developers rely on their software engineering expertise to validate and guide agent outputs, ensuring alignment with design principles and project requirements.
• Version Control: Use of version control systems helps manage multiple agents and maintain a coherent workflow.

Task Suitability of AI Agents

The study categorizes task suitability into three domains:

• Well-suited Tasks: Agents excel in generating scaffolding code, test writing, and handling repetitive tasks. Simple and well-defined tasks benefit significantly from agent assistance.
• Controversial Tasks: Mixed opinions exist on using agents for planning and architectural design. While some developers experiment with agents in high-level design, others caution against potential misalignments in strategic decisions.
• Unsuitable Tasks: Agents are deemed inadequate for tasks requiring intricate domain knowledge or complex decision-making. The lack of critical judgment and precise execution limits their use in business logic or high-stakes scenarios.

  Figure 1: Distribution of tasks and development experience. The sums exceed 99 (total # of analyzed responses) as task and experience domains may fall under multiple categories.

Developer Sentiments Toward AI Agents

Developers express a generally positive sentiment toward AI-assisted coding, citing enhanced productivity and enjoyment in the development process. The study notes that while confidence in agent utility is high, developers emphasize the necessity of human oversight to ensure quality output. The future of agentic coding is viewed optimistically, with expectations of broader adoption and integration in typical software development workflows.

Figure 2: Use of agentic coding tools in the survey. Numbers sum up to more than 99 as one user may report more than one tool use.

Conclusion

The findings from the study illustrate that professional developers use AI agents not to replace traditional coding practices but to augment their productivity and capabilities under controlled conditions. The research underscores the importance of maintaining human expertise and decision-making, especially with complex and strategic software tasks. Future developments in AI interfaces and enhanced agent capabilities could further solidify the role of AI in software engineering.