GP-SOP-007 | Planning and Operations
Decide Whether a Task Should Be Automated
A defensible decision to automate, assist, standardize, or leave the task manual.
When to use this SOP
Use this before building an automation simply because a task is annoying or technically possible.
What you need
- A clearly named task
- How often it happens
- Current time and error cost
- Systems and data involved
- Risk if the output is wrong
The procedure
Follow these steps
- 01
Describe the task from trigger to finished result without mentioning a tool.
- 02
Measure frequency, average time, avoidable errors, and the cost of delay.
- 03
Identify judgment calls, exceptions, sensitive data, approvals, and irreversible actions.
- 04
Standardize the manual process first. A moving process is a poor automation target.
- 05
Choose the smallest useful intervention: checklist, template, AI assistance, partial automation, or full automation.
- 06
Estimate build time, maintenance burden, software cost, and the time required to recover from failure.
- 07
Define a human checkpoint and a fallback manual path.
- 08
Run a limited test and compare the result with the original process.
Human checkpoint
Stop and review before continuing
Automate stable repetition. Assist variable judgment. Keep high-risk final decisions human.
Definition of done
- The current process is understood
- Benefits and maintenance costs are estimated
- Risks and exceptions are visible
- The chosen level of automation is proportionate
- A rollback or manual fallback exists
When the process gets stuck
If nobody can explain the manual process consistently, document and test it before choosing software.
Where automation fits
This assessment should happen before opening n8n. The result may still be a checklist or template rather than a workflow.
Optional AI assist
Use this after you collect the real inputs
This prompt can organize a first pass. Review the result against the SOP before using it.
Evaluate whether this task should be standardized, AI-assisted, partially automated, fully automated, or left manual. Consider: - Frequency and time cost - Input consistency - Judgment and exception rate - Data sensitivity - Cost of an incorrect result - Tool and maintenance cost - Human review point - Manual fallback Recommend the smallest useful intervention and explain the tradeoffs. Do not assume automation is the best answer. Task details: [PASTE DETAILS]