What three problems come from administering systems by hand with checklists and memorised routines?

• Error-prone: a person can skip a step or run one by mistake.
• Little evidence: hard to prove the steps were done correctly or had the intended result.
• Inconsistency: servers managed independently drift apart in their configuration.

For many years system administration meant clicking through GUIs or typing commands, guided by a checklist, some documentation or routine. Each of the three problems follows from that:

1. Error-proneness. Humans get tired and distracted. On the 40th server, step 7 gets skipped.
2. Limited evidence. Afterwards nobody can say with certainty what was done where. An audit or a post-mortem has little to go on.
3. Inconsistency. Two admins solve the same task slightly differently, and small one-off fixes pile up. This is called configuration drift: servers that should be identical no longer are, which makes maintenance hard and causes errors and instability that appear on one machine but not another.

Automation addresses all three: the same code runs the same steps everywhere, and the code plus its run logs is the evidence.

Go deeper:

• Snowflake Server (Martin Fowler) — the classic essay on how hand-tweaked servers become unique, unreproducible and fragile.
• What is configuration management? (Red Hat) — explains configuration drift and drift analysis in plain terms.