ทันเอไอ

Linear Shares Real Data from Development Teams: AI Boosts Code Output but Doesn't Cut Working Time

An analysis of real usage data from Linear reveals AI usage patterns in software teams: quantitative output clearly surges, but it comes with new kinds of workload and long-term quality concerns.

Linear, a project management platform for software development teams, has published an analysis of real usage data from its own user base to see how AI is changing the way software teams work. The clearest finding is that teams using Coding Agents (automated coding assistants) saw a significant increase in Pull Requests — for example, jumping from 21 to 65 per week in some groups.

The data also shows that roles within organizations are shifting. Non-engineers can now write more code with AI assistance, while many senior executives are returning to hands-on technical work themselves through these tools.

However, the most interesting point is that even though quantitative output has increased, overall product development time has not decreased, because AI has become a "new layer" of added work — such as reviewing and quality-controlling AI-generated code. There are also concerns that AI lacks the ability to manage system architecture complexity, which could create hard-to-manage technical debt in the long run.

This study aligns with reports from other sources. McKinsey (2025) noted that organizations getting good results from AI are those that completely overhaul their workflows and roles — not merely bolt the tools on. Meanwhile, the DORA report warns that AI adoption may negatively affect delivery stability, because code generated too quickly makes each batch of work larger and harder to review.

On the developer community side, the thread on Hacker News, which earned 178 points and more than 110 comments, mostly emphasized that "increased activity" does not guarantee real business returns, and warned that engineers may lose deep analytical thinking skills if they rely too heavily on AI.