AI-assisted engineering: Q1 impact report

DX works with hundreds of engineering organizations to measure and improve developer productivity. Through this work, we have a unique vantage point into how teams are adopting AI and what results they’re actually seeing.

This Q1 2026 AI impact report, the second in the series, covers the period of time from October of 2025 to the end of January 2026. During this time, the industry has witnessed a quarter defined by rapid model improvements, agentic orchestration, and novel, creative workflows. Analyzing data from more than 400 companies, this report provides a data-driven snapshot of AI’s tangible impact on engineering organizations.

With industry AI adoption hitting 93%, comparing AI users to non-users is no longer a viable benchmark. Our data continues to show that there is no “average” experience, AI accelerates velocity and quality for some teams, while degrading maintainability for others. Leaders can use this guide to evaluate longitudinal trends across cohorts of users and teams to understand how integration of AI into workflows can influence productivity outcomes.

Key highlights from Q1 2026

• Quality is volatile: Nearly 30% of all merged code is AI-generated. While throughput is up, some teams face a 50% increase in defects, making rigorous automated testing essential.
• Shadow AI persists: Developers frequently bypass official channels to use unapproved tools. Organizations must establish clear acceptable use policies rather than attempting to outright block experimentation.
• Developer roles are expanding: Engineering managers using AI daily now ship 4x as much code as non-users. Furthermore, 75% of designers and product managers are utilizing AI coding tools to accelerate handoffs.
• AI continues to accelerate engineer onboarding: As agents and coding assistants are increasingly used in onboarding tasks, we’ve seen new engineers start producing value within 33 days, down from the 39 day figure in the Q4 2025 report
• Agentic tools and Rust pull ahead: Agentic tools like Cursor saw a 46% increase in PR throughput for daily users. Language-wise, Rust has pressed forward, with Rust-specific reasoning improvements appearing in models such as Claude 3.7 and Gemini 3.0.

While top AI users save nearly 5 hours weekly, non-AI bottlenecks still eclipse these gains. Inefficiencies like CI wait times and meeting bloat consume vast amounts of time. AI is a powerful local optimizer, but true AI readiness requires fixing the systemic processes surrounding the code. The most effective leaders should continue to use benchmarks to inform their strategy, building on external insights while grounding decisions in data from their own organization.