AI Can't Replace Senior Engineers: Microsoft Outage Reveals Limits of AI in Complex Systems

A senior software engineer with experience building large-scale projects from scratch at companies in the United States, Europe, and Korea — both before and after the advent of AI — shares perspective on the recent Microsoft outage.

AI excels at writing code, developing applications, testing, and documentation, but the author argues this operates at a low level of software engineering. The real challenge lies in deeply understanding complex systems.

No matter how well a system is designed, unpredictable edge cases inevitably cause failures. This inherent complexity is why IT salaries far exceed those in other fields.

Systems are composed of intertwined components — databases, infrastructure, applications, and networks — with thousands of potential failure checkpoints. AI struggles here because it lacks an understanding of these complex, often nested systems.

Even rolling back a change requires pinpointing the exact root cause. In the Microsoft incident, it took a full 15 hours to diagnose the problem, potentially costing more than the combined salaries of all laid-off employees.

Senior engineers who built the systems hold the core knowledge of how they work. The author notes that a system they built could be understood in an hour, whereas an outsider relying on AI might need a day.

Ultimately, AI's best role is as an engineer's assistant, not as the developer. Unfortunately, companies often realize this only after damage has occurred.