Free report·
This is a test!
40% of developers using AI coding tools use them for editing or modifying existing code, the most common task.
This report examines how developers use AI coding tools across their day-to-day work, drawing on responses from developers who use such tools. It documents which software development tasks developers most often perform with AI assistance, how much of their coding work these tools support, and how much time they estimate they save per week across specific tools including Claude Code, OpenAI Codex, Gemini Code Assist, Amazon Q Developer, Windsurf, GitHub Copilot, Replit, Cursor, GitLab Duo and JetBrains AI. It also looks at barriers preventing professional developers from integrating generative AI into their applications, broken down by industry such as data analytics and BI, financial services and banking, and manufacturing. The findings establish that developers readily adopt AI for editing, debugging and code completion, while remaining more hesitant about tasks such as reviewing code and pull requests. The report provides benchmark figures for task priorities, the share of coding work assisted by AI, and time savings by tool.
/Key findings
- 40% of AI coding tool users cite editing/modifying existing code as a most important task, followed by debugging at 38% and code completion at 37%.
- 43% of developers using AI coding tools have very little (<25%) of their coding work assisted by these tools, while 5% have nearly all (75-100%) assisted.
- Among professional developers not integrating generative AI, 28% in data analytics and BI cite data privacy or security concerns as a barrier.
- For Cursor users, 41% estimate saving 3-5 hours per week, the highest share in that band among the tools measured.
- Reviewing code and PRs is a most important AI task for only 31% of users, lower than editing, debugging and code completion.
/Questions this report answers
What software development tasks do developers most use AI coding tools for?
Editing or modifying existing code leads at 40%, followed by debugging (38%), code completion or next edit suggestions (37%) and detecting and fixing security issues (35%).
How much of developers' coding work is assisted by AI coding tools?
43% of users report very little (<25%) of their work is assisted, 27% report some (25-49%), 26% most (50-74%) and 5% nearly all (75-100%).
What are the main barriers to integrating generative AI into applications?
Barriers include data privacy or security concerns, integration complexity with existing systems, and unclear return on investment, with data privacy or security cited by 28% of non-integrating developers in data analytics and BI.
How much time do developers save using AI coding tools?
Savings vary by tool; for example, 41% of Cursor users and 35% of Claude Code and GitHub Copilot users estimate saving 3-5 hours per week.
Are developers hesitant to use AI tools for reviewing code and pull requests?
Yes, reviewing code and PRs is a most important AI task for 31% of users, below editing, debugging and code completion, indicating more hesitancy in this area.
/Who it's for
- Developer tooling teams benchmarking AI coding tool adoption
- Product managers assessing where developers apply AI assistance
- Technology strategists evaluating barriers to generative AI integration