Microsoft AI coding agents boost pull requests by 24% in early 2026

Microsoft’s internal study of its early‑2026 rollout of Claude Code and GitHub Copilot CLI found that engineers who adopted the command‑line AI agents merged roughly 24 % more pull requests than they would have otherwise【1】. The lift persisted across a four‑month window, signaling a measurable productivity impact that goes beyond novelty usage.

Adoption dynamics and retention

The study tracked usage among tens of thousands of Microsoft engineers. Initial adoption spread mainly through social networks, meaning developers were more likely to try the tools after seeing peers use them. Retention, however, correlated more strongly with individual coding activity than with demographic factors, indicating that heavy coders kept the agents active longer. The 24 % increase in merged pull requests serves as the study’s proxy for output, acknowledging that a merged PR is not a direct measure of value but a tangible indicator of workflow impact.

Competitive implications

These findings suggest that command‑line AI agents are neither uniformly adopted nor merely a short‑lived novelty. For organizations considering similar rollouts, the data imply that visible peer usage should be a central component of deployment strategy. Competitors offering comparable CLI tools will need to demonstrate not just adoption spikes but sustained usage tied to developers’ core coding work to achieve comparable productivity gains.

What to watch

• Microsoft internal rollout milestones – further telemetry on whether the PR lift sustains beyond the four‑month window.
• Competitor responses – announcements of new CLI‑based AI coding agents or enhancements to existing tools that aim to replicate or exceed the 24 % PR lift.
• Enterprise adoption metrics – broader industry studies that compare Microsoft’s results with other large‑scale deployments of AI coding agents.

The 24 % pull‑request lift underscores that AI‑assisted coding can meaningfully accelerate software delivery, but the effect hinges on peer influence and sustained usage among active developers. Whether similar gains can be replicated at scale across other firms remains an open question.