Why AI hasn’t replaced software engineers, and won’t

2026-06-14 · Source: Simon Willison's Weblog · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning · Depth: Fundamental Awareness, quick

Summary

Arvind Narayanan and Sayash Kappor analyze why artificial intelligence has not replaced software engineers and argue against the narrative of mass AI-driven job losses, even in a sector highly susceptible to AI disruption. They highlight that current data does not support widespread AI-induced unemployment, citing New York's WARN Act filings from March 2025, where over 160 companies filed notices but none checked the AI disclosure box in the first year. The authors contend that while AI accelerates code typing, it fails to address the true bottlenecks in software engineering. These critical areas include deciding and specifying project requirements, verifying and ensuring accountability for delivered solutions, and the deep human understanding of the codebase, business context, and operational environment necessary for these tasks.

Key takeaway

For Directors of AI/ML evaluating team structures, recognize that AI tools enhance productivity but do not automate the critical human elements of software engineering. Your teams should prioritize developing deep understanding of business problems and system contexts, as these remain the primary drivers of value and are resistant to current AI capabilities. Focus your AI integration efforts on augmenting specification, verification, and debugging processes, rather than expecting full task replacement.

Key insights

AI won't cause mass software engineering layoffs because core tasks require deep human understanding beyond code generation.

Principles

• AI accelerates coding, not core engineering.
• Human understanding drives value in software.
• Bottlenecks are specification, verification, accountability.

Method

The article identifies bottlenecks through qualitative analysis of software engineers' understanding of tasks resistant to automation.

In practice

• Focus on problem definition and solution verification.
• Cultivate deep understanding of business and code.

Topics

• AI Impact
• Software Engineering
• Job Displacement
• Workforce Planning
• Automation Bottlenecks
• Human-AI Collaboration

Best for: Software Engineer, Director of AI/ML, Consultant

Related on AIssential

• Why AI hasn’t replaced software engineers, and won’t — AI primarily automates the "execute" layer of software development, leaving human "decide" and "deliver" functions intact.
• The “End of Coding” is a Myth. The Reality is Much Harder — AI shifts the bottleneck from code generation to ensuring quality, trust, and business alignment in complex systems.
• AI Isn’t Replacing Programmers — It’s Compressing Teams — AI increases individual developer leverage, compressing team sizes and shifting focus from code execution to judgment and problem-solving.
• AI’s impact on software engineers in 2026: key trends, Part 2 — AI amplifies existing engineering culture and individual skill levels, making scaled adoption complex and uneven.
• AI Is a Tool — Not a Replacement for Humanity — AI is a powerful tool whose value depends on human intent, offering an opportunity to elevate humanity, not replace it.
• How AI Is Reshaping the Way Enterprises Build Software - With Tim Sears of HTEC — AI's primary impact on software development and business will be as a catalyst for teamwork, not individual productivity.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.