When is automation worth it - a 3-number test any owner can run

Automation is worth it when a task costs you more in a year than it costs to build and keep running the thing that does it. You can check that in about ten minutes, using three numbers you already know: how many times a week the task happens, how many minutes it takes each time, and what an hour of the person doing it really costs. Multiply those into a yearly cost, put it against the honest price to build and run the automation, and you have your answer for most tasks.

Two things bend that answer, and almost every guide skips them. A task where one mistake reaches a client is worth automating earlier than the raw hours suggest. A task whose steps change every month is a bad candidate no matter how often it runs.

What three numbers do you need?#

Three, and you can pull all of them from memory or a quick look at your calendar.

1. How many times a week the task happens. Count every time, across every client or account. A report you run for ten clients happens ten times a week, not once.
2. How many minutes it takes each time. Time yourself once if you are not sure. Include the small stuff: opening the tools, copying the numbers, checking your own work.
3. What an hour of the person doing it really costs. A person paid thirty dollars an hour costs more than that once you count the higher-value work they are not doing while they copy numbers between spreadsheets.

Multiply the first two numbers to get minutes a week, turn that into hours, then multiply by about fifty weeks and by the hourly cost. That yearly number is what the task costs you today, even though it never shows up as a bill.

How do you run the math?#

Here is the test filled in for two real-shaped tasks: one that clearly passes and one that clearly fails. Run your own task down the same rows.

The reporting task eats around seven thousand dollars a year. A small single-workflow build like that usually runs a few thousand dollars, so it pays for itself inside a few months and keeps paying every year after. The quarterly task costs under two hundred dollars a year, so no build price makes sense. Do it by hand and move on.

What are the two corrections everyone skips?#

The raw hours are only half the picture. Two things move the tipping point, and the balance below shows how.

The first correction is error cost. The hours you save are only part of the value. The other part is the mistakes the automation removes. If one wrong number in this task reaches a client - a bad invoice, a missed renewal, a report with last month's figures - the cost of that one mistake belongs in the math too. A task with a high error cost is worth automating earlier than its hours alone say, because you are buying fewer mistakes, not just fewer hours.

The second correction is change rate: how often the process itself changes. Not the data, the steps. If the way you do this task is different this quarter than last, an automation built for the old way is wrong the day the way changes, and someone has to rebuild it.

When is the answer no?#

The most useful thing this test gives you is a no you can defend. A task can be annoying, repetitive, and still not worth automating, because the volume is not there or the process will not hold still. Being annoyed is not the same as it costing you money.

If the number does clear the bar but the task is simple and stays inside your own team, you may not need an agency at all. Whether to build it yourself or hire someone is the next question after this one. And when you run this test across every repeated task you own, the yearly numbers also tell you the order to work in: the biggest ones first. That order matters more than most owners think when you are trying to scale without hiring.

How do we run this test on the first call?#

We run this exact test on the first call, before anyone talks price. We ask for the three numbers, do the multiplication out loud, and add the two corrections. Often the answer is that you should not automate this yet, because the volume is not there or the process is still changing. We say so. It costs us that build, and it is why people come back later with the workflow that clears the bar.

When it does clear the bar, the first step is small on purpose. We write the task down as a one-page process map: every step a person does today, including the ones they do without thinking. Then we scope the smallest first version: one workflow, wired into the tools that already hold the data (the CRM, the project tracker, the reporting spreadsheet). It has a clear point where a person still approves the output before anything goes out. Before it runs live, we shadow it against your real history from the last month. The automation does the job on work your team already did, and we compare what it would have done to what actually happened. That is where the missing edge cases show up, before they can cost you a client. You keep the process map, the logins, and the record of every call the automation makes.

The short version: three numbers give you the hours, two corrections give you the truth, and a task that changes every month never clears the bar. Most tasks answer themselves once the math is on paper.