Should I automate this task, delegate it to AI, or just stop doing it?

How to read the four options

The table runs from cheap to costly, top to bottom. That is the whole trick.

A deleted task costs nothing and never comes back. An automated task costs one setup and then runs for free. A delegated task costs a brief once and a short review every time. A task you do yourself costs you, the scarcest thing in your business, on every single repeat.

So you want each task to fall out at the cheapest row it can. Ask the four questions in order, top to bottom, and stop at the first yes.

Most owners run it the other way. They treat every task as a Do until proven otherwise. That is how the pile stays on your desk. Flip it. A task has to earn its place there.

The order matters more than any single call

Say you skip the first question. You take the weekly report nobody reads and build a smart system that writes it for you. Now it is worse. It still wastes the reader's time, it cost you a setup, and it is harder to kill because it looks like a system.

That is the most costly mistake in the chapter, and it feels like progress while you make it. Faster waste is still waste.

One more rule makes the calls quicker. If you write "well, it depends on the client," look hard at the "depends." Is it your judgment, or a rule you never wrote down? Often it is a rule. That makes it a Delegate with a good brief, not a Do.

When should you not automate a task?

Most advice says automate anything that repeats. The book adds a second question before you build: what happens the day this rule fires wrong? Rank each task by how often it happens and by what a mistake would cost.

Two plain tests help you sort the bottom-right box. Could you write down how you make the call, so a stranger would do it the same way? And can you name the exact harm a rule would cause? If you can write it down, it is a procedure, and procedures can be automated. If you cannot name a harm, you may just like doing it yourself.

The book's sharpest example: wiring the final gallery delivery to send on its own, and one day it ships the wrong couple's photos with your name on it. A final send like that earns a human check, even when the rest of the process runs by itself.

Some tasks are three tasks wearing one coat

A monthly newsletter does not fit one row. It is three jobs. Deciding what is worth saying is a Do. Drafting it is a Delegate. Sending it on a schedule is an Automate. Cut a task at its seams first, then sort the pieces.

And some tasks fit the Do row today only because no brief exists yet. Mark those "Delegate, pending." They are yours until you write the instructions, not forever.

The reverse case matters too. A refund request has a real human call inside it. Do not automate the call. Automate everything around it, and leave the decision as a clean, fast Do.

Sunday night, a list of twelve

Nadia, the wedding photographer in the book, sat at the kitchen table with twelve tasks she had marked as habit. Twelve things to fix, and seven months of albums still owed. The list had named the problem. It had not shrunk it.

So she ran each one down the four rows. One pass, no agonizing. Two went to Delete. Four went to Automate. Three went to Delegate. Three stayed as a Do: editing the gallery, building the shot list, and deciding if a couple was the right fit.

She killed the mood board that night in about four minutes. Then she did the step most people skip. She deleted the reminder that told her to build the next one. A delete is not done until the trigger is gone, or the task crawls back within a month.

It cost her zero bookings and gave back an afternoon a week.

I like this chapter best of all because the biggest win in it uses no tech at all.

Where The Leverage Stack takes this next

This is the problem The Leverage Stack was written for.

Chapter 2, The Inventory, gives you the Task Triage Sheet, more before-and-after cases, and the four ways people sort the pile wrong, including cutting a quiet task that was actually holding things up.

Chapter 1, The Bottleneck, builds the list you sort here. Chapter 8, The Human Layer, guards the Do row, so it stays small but never empty. If you fear handing off too much, read what work should stay yours.

Why bother sorting at all? The NBER study of AI at work found AI help lifted the least experienced workers the most, up to 34% per hour. And the Census Bureau shows how many of us now run with no staff to hand work to. The triage sheet is in the free workbook.