The 3-question audit before you automate anything

I’ve audit and installed several workflows and automations across the past 15 years (automations are not new! integration and autonomy through Ai is). The most expensive AI workflows are the ones that automate the wrong thing. And the wrong thing usually looks like the right thing. Let me explain.

We automate fast. We see a task that takes time, we open Claude, we build a workflow, we save 30 minutes a week. Win, right? Not always. Half the workflows people build don’t survive the question should this task even exist.

Automation isn’t free. And this is the misconception. Every workflow you build is something you now have to maintain, monitor, and trust. If you automated the wrong thing, you’ve just made it harder to stop doing the thing you didn’t need to do in the first place.

Before you build anything, run the audit.

Three questions. Ask them about every task before you automate. If a task fails any one of them, don’t automate. Eliminate or hold.

Step 1. List 5–10 tasks you’re considering handing to AI. Drafting emails. Summarizing meetings. Generating captions. Scheduling. Research. Write them down. The list is the audit’s container.

Step 2. For each task, ask:

Question 1 — Would I do this task if it took me four times longer?

If no, you don’t have a task you want to automate. You have a task you want to eliminate. Automating a thing you secretly resent doesn’t honor your time. It just makes the resentment quieter.

Question 2 — Does someone actually need this output, or did I assume they did?

Most “obligations” are inherited. Recurring reports nobody reads. Status updates from the era of a previous client. Captions on every post because someone said you should. Before you automate the output, find out if anyone is actually waiting for it. Half the time, nobody is.

Question 3 — If I automate this, what shows up in the space it leaves?

his is the hardest question. The conscious one. The brave one.

You’re not just saving time when you automate. You’re creating space. And space doesn’t stay empty — it fills with something. If you don’t know what you want it to fill with, it will fill with more work. More tabs. More noise. More of the same.

Naming what the space is for is the practice. Deep work. Time with my kids. A walk. A pause before the next decision. If you can’t name what the space is for, you’re not ready to automate yet.

Step 3. Sort your tasks into three buckets:

• AUTOMATE — passes all three questions. Build the workflow.
• ELIMINATE — fails question 1 or 2. Stop doing it. Don’t automate it.
• HOLD — passes 1 and 2 but you can’t answer 3. Sit with it for a week. Don’t build anything yet.

Step 4. Build only the AUTOMATE list. Leave HOLD alone until question 3 is clear.

Want a partner for this? Open Claude and paste this:

I want you to help me audit a list of tasks I'm thinking about automating with AI. I'll paste the list. For each task, ask me the 3 audit questions, one at a time. Don't move on until I answer all three honestly.

After I've answered for every task, help me sort them into AUTOMATE, ELIMINATE, and HOLD. Push back if I'm being soft on question 3 — I tend to want to automate even when I can't name what the space is for.

The hardest thing about question 3 isn’t answering it. It’s that the answer often tells you something you didn’t want to know.

Sometimes the space you want to create doesn’t exist yet because you haven’t given yourself permission to want it. Maybe you can’t say I want time to be bored, to walk, to sit with my kids because part of you still believes the business requires every hour. Maybe you’ve been so busy you forgot what you’d choose if you had it.

That’s not a failure of the audit. That’s the audit working.

You catch yourself opening Claude to build a workflow before you’ve asked what the workflow is for.