In AI hyperscale operations, procedures are only useful when they control the work and the risk around the work. The site does not become reliable because it has documents. It becomes reliable when those documents define who acts, when they act, what the hold points are, and how the team verifies the result.
Inspect
Collect telemetry, verify the asset state, and confirm whether the task is routine, planned, or abnormal.
Control change
Use the SOP for routine work, the MOP for a planned state change, and the approval path that matches the risk.
Stabilize
Use the EOP to contain an abnormal event, communicate fast, and restore the site to a known-safe condition.
What each document actually does
- SOP: covers repeatable work that should produce the same result every time.
- MOP: covers a controlled state change, such as switching, isolation, transfer, or maintenance that can affect the operating environment.
- EOP: covers abnormal conditions where the priority is stabilization, escalation, and safe recovery.
The strongest sites treat those three documents as one operating system. The SOP tells the team what normal looks like. The MOP defines the boundary before change. The EOP gives the team a clear response when the system no longer behaves normally.
What a usable procedure must include
- Scope and asset identification.
- Prerequisites and safety checks.
- Step-by-step actions with clear hold points.
- Expected outcome after each critical step.
- Rollback, backout, or escalation path.
- Sign-off, timestamps, and evidence capture.
If any of those pieces are missing, the procedure becomes harder to trust and easier to improvise. In a hyperscale facility, improvisation is usually where small issues become large ones.
Data Center Health 