Closable or open?

The idea

Some tasks AI can finish. Others it can only help you think about. Two questions tell them apart:

1. Can I write down what “correct” means, well enough to check the result?
2. Does anyone get to move after I act: a customer, a competitor, the other side of a deal, a reader who reacts?

If you can check it and nobody moves after you, the task is closable. Hand it over and check the result against what you wrote down. If either answer goes the other way, it's open. Use AI for options, drafts and arguments, and keep the decision yourself.

Why it works

A closable task has a finish line you can see, so you can tell when the AI got there. An open task has no fixed finish line: the other side reacts, and what counted as right changes once you move. AI's answer to an open problem still looks finished, and that's the trap. A polished pricing plan invites you to stop thinking at exactly the point where your judgement was the job.

“Hard” and “open” are different things. Reconciling a year of messy receipts is hard but closable. Replying to one upset customer is short but open.

How to do it

1. List the tasks from a normal week. Ten is plenty.
2. Ask the two questions of each one, honestly.
3. For closable tasks, write the check first: “every receipt matched or listed as missing”. Then hand it over and check the result against it.
4. For open tasks, ask for options and arguments, not answers: three approaches, the strongest case against each, what the other side might do next.
5. Add a third question for anything open: can I undo it? If not, see one-way doors.

The closable-or-open sorter walks you through this on the page.

Here is a list of tasks from my week. For each one, answer two questions:
1. Could I write down what a correct result looks like, clearly enough to check it?
2. Does anyone (a customer, competitor, the other side, an audience) get to react after I act, in a way that changes what's right?

Then sort them into CLOSABLE (I can hand it over and check it) and OPEN (I should use you for options and decide myself). For each closable task, suggest the check I should write. For each open task, suggest what to ask you for instead of an answer.

My tasks:
[list]

Worked example

Common mistakes

• Calling something closable because the AI's answer looks finished. Ask whether you could check it.
• Handing over a closable task without writing the check. Then you're just reading and hoping.
• Asking an open question for “the answer”. Ask for options and the case against each.
• Treating a clean result on the easy part as proof the hard part is fine.