What should a small business automate first? A three-question test

Most owners we talk to do not have an automation problem. They have a "where do I start" problem. There are ten things that eat time every week, every one of them looks automatable, and picking the wrong one first is how a project stalls out and sours the whole team on the idea.

So before you buy another tool or ask anyone to build anything, run the candidate workflow through three questions. If it fails any of them, it is not the one to start with.

Question 1: Does it happen the same way every time?

Automation is good at repetition and bad at judgment. A task that follows the same steps in the same order most of the time is a strong candidate. A task where the right move depends on reading the situation is not, at least not yet.

Take a common example. A 20-person business sends a welcome packet to every new customer. The steps are identical each time: pull their details, fill the template, attach the two PDFs, send. That is a clean fit. Compare that to deciding which customers get a discount, where the answer depends on history, relationship, and gut. Same department, very different readiness.

If you cannot describe the steps without saying "it depends" more than once, pick a different workflow to start.

Question 2: Is the data already sitting somewhere usable?

A workflow can only act on information it can reach. If the details it needs live in one person's head, or in a paper folder, or scattered across text messages, the automation has nothing to stand on.

Ask where the inputs actually live. A form, a spreadsheet, a CRM field, or an email with a predictable shape all count as usable. A "Karen just knows who the good vendors are" does not count until Karen writes it down. Sometimes the real first project is not automation at all. It is getting the data into one place a system can read.

Question 3: Who reviews the output before it counts?

This is the one people skip, and it is the one that keeps you out of trouble. Every useful small-business automation we build has a named human who checks the result before it goes to a customer or hits the books. The software drafts. A person decides.

If you cannot name the person who will own the review, the workflow is not ready. Not because the tech cannot do it, but because nobody has agreed to catch the mistakes it will make. Drafts, not decisions. That line is the whole job.

Putting it together

Here is the test in one pass. Say you are weighing whether to automate your overdue-invoice follow-up.

1. Does it happen the same way every time? Mostly yes. A reminder goes out on a schedule until the invoice is paid.
2. Is the data usable? Yes, if your invoices live in software that knows what is paid and what is not.
3. Who reviews the output? The bookkeeper approves the list each morning before anything sends, so nobody gets chased for a bill they already paid.

Three yeses. That is a workflow ready to build. Now compare it to "automate our pricing decisions," which fails the first question and the third. The test does its job by telling you which one to leave alone.

What to do Monday

Write down the two workflows that cost your team the most time this week. Run each through the three questions above. The one with three clear yeses is where to start. If both fail on the data question, your first project is getting that data into one place, and that is normal.

If you want a structured version of this, the free AI Readiness Checklist walks your business through the same logic and gives you a 0 to 10 score in about 15 minutes. It will not sell you anything. It will tell you whether the first workflow is sitting in front of you or one step of prep away.