Karibunest

You have a list of things you could automate. Most people do. Fourteen items, written down over a few months, each of them genuinely true, and none of them started.

The list is not the problem. The absence of a ranking is.

Ask most people which task they would automate first and they name the one that irritates them most. That feels like the obvious answer and it is usually wrong, because irritation has almost no relationship to cost.

The task that irritates you might happen twice a year. The one quietly consuming four hours a month might not irritate you at all, because you have stopped noticing it.

Why the obvious ranking fails

Three things make a task expensive, and only one of them is unpleasantness.

Frequency. A task done weekly costs fifty-two times its duration each year. A task done quarterly costs four. Two tasks that feel identical in the moment can differ by an order of magnitude across a year.

Duration. The actual clock time, including the part before you start and the part after you finish.

Variability. Whether the task is the same every time or different every time. This is the one people leave out, and it decides whether automation is even possible.

Annoyance correlates with none of these reliably. It correlates with how the task feels while you are doing it, which is a fact about you rather than about the task.

The Three-Number Test

Take any candidate task and write down three numbers. It takes about two minutes per task and it will reorder your list.

1. How many times a month?

Count occurrences, not intentions. If you are unsure, look at the evidence rather than your memory: sent messages, calendar entries, files modified, the folder it lives in.

People overestimate frequency for tasks they dislike and underestimate it for tasks that have become automatic. Both errors are large.

2. How many minutes each time?

Time it once with a clock rather than estimating. Include everything. Opening the tool, finding the previous version, the interruption you cause yourself halfway through, and filing the result.

Most people find the real number is between one and a half and three times their estimate. The gap is in the parts either side of the work itself.

3. How similar is each instance, out of ten?

Ten means identical every time. One means it is a different problem wearing the same name.

This number is the gate. Below about six, automation handles the common case and quietly mishandles the rest. That is worse than doing it by hand.

Multiply and rank

Frequency times duration gives you minutes per month. Multiply by twelve for the annual cost.

Then sort by annual cost, and discard anything scoring below six on similarity regardless of how expensive it is.

What survives is usually not what you expected. The task at the top is frequently something small and dull that happens often, rather than the large irritating thing that happens occasionally.

A worked example

Here is a real list, scored.

The monthly report has the highest annual cost at thirty-six hours, and it scores four on similarity. Every month is genuinely different. Automating it would produce a plausible document that is wrong in ways nobody catches, which is the most expensive possible outcome.

The winner is the weekly status update: twenty hours a year, and a nine on similarity because the structure never changes, only the content.

That task was fifth on the original list, ordered by annoyance.

What to do with the result

Start with the top survivor and nothing else. A single task properly handled beats four half-built ones, and the second is much cheaper than the first because you already know the method.

Before building anything, check the payback: estimate the build time, double it, and divide by the minutes saved per instance. If it does not pay back within three months, do it by hand and stop feeling bad about it.

Try it this week

1. List every repeated task you can think of. Aim for at least ten.
2. Write the three numbers for each. Use evidence for frequency, a clock for duration.
3. Multiply, rank, and strike out anything below six on similarity.
4. Take the top survivor. Ignore the rest.

You will know it worked when the task at the top is not the one you would have named before you started.

This is adapted from chapter 5 of The Automated Advantage, which covers the full method: a seven-day log for finding the tasks you have stopped noticing, the Automation Ladder for deciding what to do with each one, and a thirty-day plan.