The industry consensus is that you should automate everything. Every manual process is viewed as a legacy inefficiency waiting to be containerized. This is a fatal oversimplification. Automation does not fix a broken process; it permanently cements it into the infrastructure.
⸻
The Illusion of Efficiency
When an organization identifies a painful, convoluted deployment workflow, the reflex is to script it. We spend weeks building a CI/CD pipeline that perfectly replicates the manual absurdity. We did not remove the friction. We merely automated the friction. The underlying architecture remains deeply flawed, but because it executes quickly, we stop interrogating why it exists. Efficiency masks the rot.
⸻
The Debt Amplifier
An automated bad decision scales infinitely. If your provisioning process blindly allocates excessive memory to every microservice, manual intervention provides a natural speed limit. Once you automate it, you can drain the entire cloud budget in a weekend. The organization celebrates the velocity of the rollout while completely ignoring the velocity of the technical debt. Automating stupidity is the fastest way to orchestrate a systemic failure.
⸻
The Loss of Context
When you automate a brittle process, the institutional memory surrounding it evaporates. The script becomes a black box. The original engineers leave, and the remaining team treats the automation as a law of nature. When the script inevitably fails because an undocumented edge case emerges, no one knows how to manually intervene. The automation that was supposed to save time has now introduced catastrophic risk.
⸻
Fix the process before you write the script. If the workflow is painful, interrogate the steps instead of immediately trying to bypass them with code. Do not automate a system you do not fully understand. The survivor knows that sometimes the manual friction is the only thing preventing a larger disaster. Let it intentionally hurt.
End.