A year ago, many software teams were still deciding whether AI coding assistants were worth trying. Recent survey data suggests that question has largely been answered. Most professional developers now use AI coding agents, tools that can read a codebase, make changes across several files and run commands, as a regular part of their work.
Two recent publications give a useful picture: the JetBrains Developer Ecosystem Survey 2026, published in August, and a Stack Overflow review of its survey history, published on September 30 ahead of its full 2026 results.
What the surveys found
JetBrains: weekly use is now the norm. JetBrains surveyed more than 15,000 professional developers worldwide between May and July 2026. Its key findings:
- 90% of respondents used AI coding agents at work at least once a week, and 68% used them every day.
- Claude Code was the most widely used agent at work, at 39% (47% in the United States).
- GitHub Copilot was at 21%, down from 29% in the previous survey, and Cursor fell from 18% to 12%.
- OpenAI's Codex grew from 3% to 16%.
The headline is less about any single product and more about how quickly the market is moving. Tools that led a year ago have lost ground, and new ones have grown fast.
Stack Overflow: steady growth, with more caution. Stack Overflow's look back at its own data shows AI tool use among developers rising from 44% in 2023 to 62% in 2024 and 79% in 2025. A pulse survey in April 2026 found that use of AI agents had almost doubled, to 59%, compared with 31% in 2025.
The same review also highlights a more careful mood. Among people learning to code, favorable views of AI tools fell from about 72% in 2024 to about 53% in 2025. Developers are using these tools more, but they are also more realistic about their limits. Stack Overflow says its full 2026 results are due shortly.
Companies are investing in the skills gap. The demand for people who can put AI to work inside real businesses is also showing up outside the surveys. On October 2, Anthropic announced a $100 million program, the Claude Frontier Academy, to train 10,000 engineers by the end of 2027 in deploying AI systems inside large organizations.
Why it matters
For developers, working well with AI agents is quickly becoming a core skill, alongside testing, code review and system design. Knowing when to trust an agent's output, and when to slow down and check it, matters more than knowing every shortcut.
For business owners and clients, the change shows up in a few ways:
- Routine work such as boilerplate, tests, documentation and small fixes can often be done faster.
- Review and quality control become more important, not less. Code written quickly still needs to be correct, secure and maintainable.
- Questions about tooling are now fair to ask a software partner: which AI tools are used, where your code and data go, and how output is checked.
Survey numbers should be read with care. Each survey has its own audience and method, and adoption numbers do not measure quality or productivity directly. They do, however, show clearly where everyday practice is heading.
What we recommend
For engineering teams:
- Write down how you use AI. A short internal guide covering approved tools, what can and cannot be shared with them, and how AI-written code is reviewed avoids confusion and risk.
- Keep humans responsible for every merge. Treat agent output like a pull request from a new team member: read it, test it, and only then accept it.
- Strengthen your safety nets. Good automated tests, linting and security scanning matter more when code is produced faster.
- Avoid lock-in to one tool. The market is shifting quickly. Keep workflows simple enough that switching tools is not a major project.
For businesses hiring a software partner:
- Ask about AI policy directly. A good partner should explain which tools they use and how they protect your code and data.
- Ask about review and testing. Speed is useful only if quality controls are in place.
- Focus on outcomes. The right question is not whether AI was used, but whether the software is reliable, secure and easy to maintain.
AI coding agents have moved from experiment to everyday tool in a short time. The teams that benefit most will be the ones that pair them with clear rules and careful review.
