Microsoft's internal rollout of Claude Code and GitHub Copilot CLI produced a measurable output lift — but the story is more nuanced than a headline number.
Researchers studying tens of thousands of Microsoft engineers over a four-month window in early 2026 found that engineers who adopted CLI coding agents merged roughly 24% more pull requests than they otherwise would have. First use spread primarily through social networks — meaning engineers tried the tools because colleagues were visibly using them, not because a memo told them to. Retention correlated more closely with how much an engineer was already coding than with any demographic factor. The study used merged PRs as its output proxy, with the authors explicitly noting that a merged PR is not the same as the value it delivers.
The finding matters because organizations are being asked to make expensive bets: token spend at scale can run into millions of dollars annually, and a misread on adoption or retention turns a rollout into a cost center with no velocity gain. This study offers something rarer than a vendor benchmark — a large-sample internal audit from a company with the scale to see real signal. The 24% lift also held across the full four-month observation window, which rules out the simplest dismissal: that early enthusiasm fades.
The skeptical read is that PR volume is a convenient metric for a tool that helps write code faster, not proof that the underlying software improved. Still, for any engineering org weighing a Claude Code or Copilot CLI deployment, the peer-network adoption finding is the most actionable takeaway — budget for visibility, not just licenses.