Three questions to ask before you automate anything

The question people usually ask about a task is "can AI do this?" It is the wrong place to start, because the answer is increasingly yes, and yes tells you nothing about whether you should.

Here are three questions that take about two minutes and prevent most of the regret.

1. Does this actually require me?

Not "is this important", and not "am I good at it". The question is whether your judgment, relationship, or presence is what makes the task work.

Plenty of tasks land on our desks for reasons that have nothing to do with fitness. We inherited them. We were the only one available once. We built the spreadsheet three years ago and never handed it on. Habit is not the same as responsibility.

If the honest answer is "no, I have simply always been the one doing it", that is a strong candidate for support, delegation, simplification, or removal.

If the answer is "yes, and here is exactly why", write down that reason. You will need it for the third question.

2. If this went wrong, would I still own it?

This one separates useful assistance from quiet abdication.

Ask what happens when the output is wrong, and it will sometimes be wrong. Who notices? Who is accountable? Could you explain how the result was produced to the person it affects?

If you would still stand behind the outcome, and you have a way to check it before it matters, you are augmenting your work. If your instinct is to say "the tool did that", you have not automated a task. You have moved responsibility somewhere it cannot actually sit.

Three practical safeguards, written once per workflow, cover most of this:

1. A verification rule. What gets checked, by whom, before it counts.
2. A disclosure rule. When you tell people how the work was produced.
3. An information boundary. What never goes into the tool, no matter how convenient.

Ten minutes of writing. It makes everything downstream repeatable.

3. What is the recovered time for?

The one most people skip. If you cannot say where the saved capacity is going, it will be absorbed by more work, and the gain becomes invisible.

Finish the sentence before you build anything: the time this saves is for ____.

What this looks like in practice

A recurring report that took an hour a week:

• Does it require me? The gathering and formatting, no. The interpretation and the recommendation, yes.
• Would I own it? Yes, provided I read the numbers myself before they go anywhere. That is the verification rule.
• What is the time for? Two extra conversations a week with the people the report is about.

So the answer was not "automate the report". It was "automate the preparation, keep the reading, and spend the difference on the conversations the report was supposed to start".

That is usually how it goes. The right answer is rarely all or nothing, and it is often smaller than the tooling would suggest.

When the answer is no

Sometimes the honest recommendation is: do not automate this. Keep it human.

That is not technophobia, and it is not a lack of ambition. It is the recognition that some things carry their value in the doing, and that speeding them up removes the point.

The F.A.I.T.H. audit is built around these questions. Twenty statements, five minutes, and one clear next step.