“Prepare the workspace” is an instruction that makes sense to someone who already knows the job. A new person still needs to know what prepared looks like.
Start a checklist with observable completion conditions. In a fictional packing area, “prepare the workspace” could become: clear unrelated materials, place the approved order sheet at the station, and confirm the required packing supplies are present.
Those details must come from the real process owner. AI should organize them, not invent your operating standard.
Use this input structure:
When the task begins.
What must be available first.
The current approved steps.
What tells the operator each step is complete.
What should cause them to stop and ask for help.
Then try:
Convert these supplied instructions into a draft checklist. Preserve their order unless I explicitly authorize a change. Each step needs an action and an observable completion condition. List missing details as questions for the process owner. Do not invent safety rules, quantities, approvals, or company policy. Separate your improvement suggestions from the current procedure.
Review the draft with someone who performs the task. Give them a sample scenario and ask where they would hesitate. That hesitation is useful evidence: the instruction may still depend on knowledge that isn’t written down.
Version the approved checklist with an owner and review date. Keep the original instructions accessible so a tidy rewrite doesn’t erase a necessary detail.
The success condition is simple: another person can follow the approved version and recognize when to stop. If they still need the author beside them, the checklist needs another pass.
Try a free sample
Want a reusable starting point? Open the SOP Starter Pack and copy the free worksheet at the top. The paid section adds guidance and worked examples to help you use it.
Become a paid reader for the complete pack archive and two new implementation packs each month.
—Future Proof