Avoiding human error as cause - the systemic approach

Human error is never the root cause; it is a symptom of a system that allows the error. This lesson covers the discipline of designing systems that prevent mistakes.

The principle

Humans make mistakes. The system should:

• Prevent the mistake (validation, approval workflows).
• Catch the mistake before impact (review, testing).
• Limit the impact when the mistake happens (blast radius).
• Recover quickly when the mistake causes harm (runbook).

The principle: the system is responsible for the human’s success.

Examples

• ‘Operator ran the wrong command’ -> the system should prevent the wrong command from running.
• ‘Admin forgot to update the config’ -> the system should have an automated config update.
• ‘Deploy broke’ -> the system should have a CI/CD pipeline with tests.
• ‘Engineer missed a CVE patch’ -> the system should have automated patch management.

In each case, the human made a mistake, but the system allowed the mistake. The fix is the system, not the human.

Design for prevention

• Validation: scripts check before applying. A bad config is rejected.
• Approval workflows: changes require review before apply.
• Testing: changes are tested in staging first.
• Automation: humans do not run critical commands directly; scripts do.
• Reversibility: every change has a tested rollback.
• Blast radius: limits on what a change can affect.
• Detection: monitoring catches mistakes quickly.

Build a culture

The culture:

• Errors are learning opportunities, not blame.
• The system is responsible, not the human.
• Post-incident reviews are blameless.
• Improvements are tracked and implemented.

The culture takes time to build. Start with the principle and apply it consistently.

Knowledge check