Senior Developers Bet on Generative AI

Senior developers are increasingly integrating generative AI into their workflows, not only to generate code but also to optimize their daily processes. According to Cyril Perraud, Chief Technology Officer at Nexenture, AI acts as a "skill multiplier," allowing developers to feel "augmented." The annual Stack Overflow survey for 2025 reveals that 84% of developers are using or considering using AI tools, an increase from 76% in 2024. Furthermore, 51% of professional developers use these tools daily.

AI is primarily used to automate repetitive tasks. Guillaume Thirard, a senior developer at Ornikar, entrusts AI with creating documents for meetings and summarizing reports. He also combines the agentic coding tool Claude Code with Copilot, the latter being very powerful for quickly writing a small script or making minor modifications without needing to specify all the context. AI is also employed for prototyping and provides an overview of project architecture, as explained by François Ibert from the Webcom agency. AI is also used for debugging and software improvement, according to Fabien Dalla-Valle from Metadot.

Despite these advantages, some complex tasks remain difficult to delegate to AI. A study by Atlassian shows that 99% of developers save time thanks to AI, and 68% save more than 10 hours per week. However, a survey by METR indicates that AI can slow developers down by 19% on complex tasks. Frustrations are also present: 66% of developers find that AI solutions are "almost correct, but not quite," and 45% of Stack Overflow responses indicate that debugging AI-generated code takes the most time.

Developers often use multiple tools to meet the needs of their projects. According to Gitlab, 60% of teams use more than five development tools. Among them, Cursor is popular for its integration with code, but Claude Code is particularly cited by senior developers for brainstorming and solving complex problems.

Gstack, a new tool, is experiencing rapid success with 50,000 stars on GitHub in just a few days. It automates architecture and code review, replacing the "AI intern" with a sophisticated pipeline. However, human verification remains crucial. Developers often need to review and correct the code generated by AI, which can be verbose and lack optimization. François Ibert emphasizes that AI cannot yet replace a human developer, as it does not consider all possibilities and requires a detailed prompt to be effective.