AI isn’t making developers more productive – it’s making them busier

2026-06-11 · Source: LeadDev · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning · Depth: Intermediate, quick

Summary

A new study from the Massachusetts Institute of Technology and the Wharton School of Business, analyzing over 100,000 GitHub developers, reveals that AI coding tools are making developers busier rather than more productive. While task-level coding activity has significantly increased—with autonomous asynchronous agents boosting lines of code written by 741%—this only translates to a 20% increase in actual software releases. The research tracked the impact of successive AI tools, from early autocomplete features like GitHub Copilot to synchronous and asynchronous agents, finding that initial coding gains attenuate sharply as work moves up the organizational hierarchy. This bottleneck, termed the "weak-link hypothesis," stems from human constraints in processes like code review, integration, and release management, which cannot keep pace with the volume of AI-generated code. Consequently, the developer's role is shifting from primarily writing code to evaluating it.

Key takeaway

For engineering managers adapting to AI coding tools, recognize that your team's bottleneck has shifted from code generation to evaluation, review, and release processes. Your focus must move beyond increasing lines of code to optimizing the entire software development lifecycle. Re-evaluate your current review, testing, and maintenance workflows to prevent work-in-progress from piling up and ensure AI's benefits translate into actual software releases and user value.

Key insights

AI coding tools increase raw code output, but human-centric downstream processes become the new bottleneck.

Principles

• System throughput is limited by its slowest stage.
• Developer roles are shifting to code evaluation.
• Unadapted processes create work-in-progress piles.

In practice

• Overhaul code evaluation and testing.
• Prioritize code review as a constraint.

Topics

• AI Coding Tools
• Developer Productivity
• Software Development Lifecycle
• Code Review
• Bottleneck Analysis
• AI Agents

Best for: CTO, VP of Engineering/Data, AI Product Manager, Software Engineer, Director of AI/ML, MLOps Engineer

Related on AIssential

• Writing Code vs. Shipping Code: Productivity Effects Across Generations of AI Coding Tools — AI coding tools boost task-level velocity, but human bottlenecks limit overall software shipping productivity to a 53% maximum gain.
• I dont get the "AI will replace devs" angle — AI is rapidly transforming software development, reducing demand for junior roles and shifting the economic landscape for individual contributors.
• ‘Tokenmaxxing’ is making developers less productive than they think — Increased AI code generation correlates with disproportionately higher code churn and revision rates, undermining productivity claims.
• Coding agents are giving everyone decision fatigue — Coding agents shift software development bottlenecks from code generation to human judgment, intensifying work and causing decision fatigue.
• The AI Coding Supremacy wars — AI coding models are rapidly advancing, significantly increasing their contribution to public code commits.
• Paradox in action — Cheaper software production via AI increases demand for new applications, sustaining human software engineering roles.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by LeadDev.