Turn a client SOP into a one-page task card

Choose one repeatable SOP

Pick a procedure that runs at least weekly and has a visible output. A task that appears once a quarter will not give the assistant enough repetition to test the card. Choose one that currently causes questions or rework, because the card should solve a real problem. The first card is a pilot, not a complete rewrite of the client's documentation.

Read the full SOP once without editing it and mark the steps the assistant actually performs. A long procedure often contains background, policy, and history that the task does not need. Separate the operational steps from the context that helps someone understand why the rule exists. The card should carry the action and the reason, not the entire manual.

Extract the trigger and outcome

Write the trigger as an event the assistant can recognize, such as a new order, a returned form, or a Monday morning report request. Write the outcome as a finished result, such as a reconciled spreadsheet saved in the shared folder. Put both at the top of the card in one sentence each. If the trigger or the outcome is unclear, the rest of the card will be guesswork.

Name the person who requests the work and the person who reviews it. State whether the assistant should wait for a request or run the task on a schedule. A card without a trigger tends to produce work at the wrong time. The trigger and outcome are the two lines the assistant should read first.

List the exact steps and tools

List the steps in the order the assistant performs them, using the same tool names the client uses. Add the login or folder path where the work starts and the place where the output belongs. Keep each step to one action so a missed step is easy to spot. A step that says handle the request is not a step.

Mark the steps that require a decision, such as choosing a category or approving a small adjustment. Add the rule that governs the decision and the person to ask when the rule does not fit. Separate the mechanical steps from the judgment steps so the reviewer knows where to sample. The card should show where the work can go wrong.

Add examples and stop rules

Attach one complete example and one corrected example to the card. The complete example shows the standard, while the corrected example shows the most common mistake and the fix. Keep the examples short enough to scan during the task. A card with examples is easier to trust than a card with only instructions.

Write the stop rules as observable conditions, such as a missing document, an amount above the limit, or a request from a new contact. For each stop rule, name the owner and the channel for the question. A stop rule that says use judgment gives the assistant no help. The stop rules are the part that keeps the task inside its authority.

Test the card with another person

Ask someone who did not write the card to complete the task using only the card and the example. Watch where they pause, search for a missing detail, or ask a question. Those pauses are the gaps the writer cannot see. Record the gaps and fix the card before the next live run.

Compare the test output with the standard example and note every difference. Decide whether the difference came from the card, the tool, or the tester's judgment. Fix the card first, then the tool or the training. A test that produces a perfect output without a question may mean the tester already knew the task.

Keep the card versioned

Put a version number, a date, and an owner at the top of the card. Store the card beside the task instructions so the assistant does not have to search two systems. When a step changes, update the card and the example in the same edit. A card that changes without a version is difficult to audit and easy to contradict.

Keep the previous version available for the review cycle so a correction can be traced. Remove the old version only after the new one has been used successfully. Tell the assistant which line changed instead of asking them to reread the whole card. Version control is what makes a one-page card reliable over time.

Review the card after a month

Read the questions and corrections from the first month of use. Add the missing detail to the card and remove any step the assistant never performs. Check whether the stop rules caught the cases they were meant to catch. The card should become shorter and more precise with use.

Ask the reviewer whether the card made the work easier to sample. If the reviewer still needs the full SOP to judge the output, the card is missing a standard. Add the standard and test it on the next output. The review closes when the card can stand alone for the routine case.

Check whether the card still matches the client's current procedure before the next busy cycle. A small policy change can make the trigger or the stop rule wrong. Read the card beside the source SOP and mark any line that no longer matches. Update the card in the same edit as the procedure so the two do not drift apart.