# You bought the AI tool. Are your engineers using it? > Source: > Published: 2026-09-01 09:54:22+00:00 This is your last article that you can read this month before you need to [register](/register) a free LeadDev.com account. Estimated reading time: 3 minutes **Key takeaways:** **Claude Code leads adoption at 78%**, but daily use** drops to 50%**. One leader calls it a procurement problem disguised as an adoption problem with engineers going around what got approved to find something better.**GitHub says 88% of Copilot users feel more productive**. Only 26% of leaders surveyed for LeadDev’s report saw a real gain, and just 31% of organizations are even measuring impact.- The real bottleneck is **waiting for tests, builds, and deploys**. Since the launch of[ OpenAI’s ChatGPT](https://leaddev.com/technical-direction/meet-speaker-evan-morikawa-how-openai-scaled-chatgpt) in November 2022, generative AI tools like GitHub Copilot and[ Cursor](https://leaddev.com/technical-direction/how-jit-overcame-developer-resistance-shift-cursor) have quickly become staples in many developers’ workflows. Their widespread adoption has fueled widespread AI hype –[ even in boardrooms](https://www.computerweekly.com/news/366554314/Gartner-Execs-put-generative-AI-on-business-agenda?utm_source=chatgpt.com) – about AI’s potential to significantly boost engineering productivity. However, [LeadDev’s AI Impact Report 2026](https://leaddev.com/the-ai-impact-report-2026) found that buying an [AI tool](https://leaddev.com/ai/best-ai-coding-assistants) doesn’t mean engineers will make it part of their daily workflow. According to the report, based on a survey of nearly 600 respondents, enterprise adoption is led by [Claude Code](https://leaddev.com/ai/why-microsoft-engineers-are-using-claude-code) at 78%, followed by Claude (model/chat) at 64% and GitHub Copilot at 56%. Yet being approved and funded doesn’t guarantee daily dominance: when asked which tool sees the most active use, Claude Code drops to 50%, with Claude (model/chat) at 18% and GitHub Copilot at 14%. ## Your inbox, upgraded. Receive weekly engineering insights to level up your leadership approach. Dan Moore, senior director, CIAM strategy and identity standards at FusionAuth,, highlights an important distinction in the report: being “adopted” doesn’t necessarily mean a tool is being “paid for consistently.” “That gap is a procurement problem masquerading as an [adoption problem](https://leaddev.com/ai/ai-adoption-has-to-be-driven-from-the-top),” he adds. “The [tools engineers are gravitating toward](https://leaddev.com/ai/your-ai-coding-tools-buying-checklist-for-2026) aren’t always the ones the organization funded. When you see that pattern, the right question isn’t ‘why aren’t engineers using what we approved?’ It’s ‘why are engineers going around what we approved to find something better?’” ## The AI hype wave The tools companies buy in frequently miss the mark because they don’t truly understand the day-to-day nature of engineering tasks. According to [GitHub, ](https://github.blog/news-insights/research/research-quantifying-github-copilots-impact-on-developer-productivity-and-happiness/)88% of Copilot users feel more productive when using the coding assistant. [Anthropic’s own engineers reported](https://www.anthropic.com/research/how-ai-is-transforming-work-at-anthropic?_bhlid=9b8836c5e2ca155c2d8a3894cd74c5847cfd7aa1&utm_source=) a 50% productivity boost when using Claude, while the company saw a 67% increase in merged pull requests per engineer per day after adopting[ Claude Code](https://www.anthropic.com/research/81k-economics?_bhlid=f2d4a41daac722df4c87ebd91b627dfd936bd352&utm_source=). However, only 26% surveyed for LeadDev’s report found a significant [boost in productivity.](https://leaddev.com/velocity/productivity-isnt-always-fast) Just 31% of organizations measure the impact of AI-powered developer tools, up from 18% last year. While that’s a step forward, most are still making decisions without hard evidence. Another 48% are working out how to measure success, suggesting the tools are being adopted faster than organizations can assess their value. “The bottlenecks that we tend to see at companies are not in the hands-on keyboard time; [it] is in the time waiting for the test to pass or fail, or for a build or deploy that won’t happen for another two to three days,” says [Rebecca Murphey](https://leaddev.com/community/rebecca-murphey), field CTO of Swarmia. “There’s a considerable and genuine need for AI tools within teams, but there’s a real detachment from the[ software development lifecycle (SDLC)](https://leaddev.com/career-development/using-experiments-bring-security-your-software-development-life-cycle) they’re meant to help, and there’s still a lot to learn and distribute about how to best use AI tools,” says Andrew Zigler, senior developer advocate at LinearB. Moore echoes this sentiment. “If you can’t name the problem you’re solving, measure it, and prove the tool improved it, you’re just riding the hype wave.” ## More like this ## Ask your devs about AI tools [AI adoption starts in the boardroom](https://leaddev.com/ai/ai-adoption-has-to-be-driven-from-the-top) – but succeeds with developers. To address core challenges effectively, engineers need to be included in discussions with leadership about implementing AI tools. This collaboration helps determine the appropriate use cases for AI within the organization and the specific problems it should aim to solve. Moore bets that most AI tooling decisions are made by [engineering leaders ](https://leaddev.com/the-engineering-leadership-report-2026/)who aren’t explicitly thinking about the pipeline. “Did anyone ask the developers? If so, they would have gotten different input than a procurement evaluation aimed at increasing development velocity. I’d go so far as to say software engineers must be involved in choosing AI tools their organizations invest in.” Zigler agrees. He argues that software engineers should be closely involved in this process, particularly when it comes to identifying and explaining bottlenecks. Whether they need systems to slow down or speed up, [more tokens](https://leaddev.com/reporting/the-tokenmaxxing-hype-didnt-last-long), or access to different tools, engineers should have clear channels for raising these needs and escalating requests.