What to automate first in a small business: a template to decide

Almost every business owner I talk to wants to automate "something". Almost none of them know where to start. And when they do choose, they tend to pick whatever is most visible: a chatbot on the website, an agent that answers WhatsApp, a new tool they saw in a video. Sometimes they get it right. Often they don't, because what eats most of their time is not what makes the most noise.

This article isn't about tools. It's about deciding. I'll walk you through a simple method to compare the repetitive tasks in your business using four criteria —frequency, time, errors and ease of implementation— plus a table you can copy and fill in with your own numbers in half an hour. By the end you'll have a ranked list of what to automate first, what to leave for later and what isn't worth touching.

The most common mistake: starting with what's visible, not what's heavy

When a small business decides to automate, one of three things usually happens. First: they pick the task that annoys them most emotionally, even if it takes little time. Answering "what time do you open?" for the fifth time in a day is irritating, but it may add up to ten minutes. Second: they pick what's fashionable. An AI agent for WhatsApp sounds like the future; an automatic appointment reminder sounds boring. Third: they pick what the vendor knows how to sell, not what the business actually needs.

The outcome is similar in all three cases: money and energy go into something that barely improves day-to-day work, and the feeling left behind is that "automation wasn't for us". That's not true. They simply started with the wrong task.

The alternative is boring, but it works: take inventory, measure and score. You don't need any software for this. A notebook or a spreadsheet is enough.

The four criteria for comparing tasks

Every repetitive task in your business can be described with four questions. To compare them with each other, each one is scored from 1 to 5.

1. Frequency: how often does it happen?

A task done once a year doesn't deserve the same effort as one repeated twenty times a day. Frequency multiplies everything else: every minute you save on a daily task is saved more than two hundred times a year.

2. Time: how long does each instance take?

Here you measure the time of a single repetition, start to finish. Not just the moment you type the message: also finding the customer's number, opening the calendar, checking the details and getting back to what you were doing. That "getting back to what you were doing" is the cost almost everyone forgets to write down.

3. Errors: what happens when it goes wrong?

Not every task carries the same risk. If you forget to post on Instagram one Tuesday, nothing serious happens. If you forget to remind someone about an appointment and the slot stays empty, you lose money. If you get a figure wrong on an invoice, you may end up with a tax problem. This criterion doesn't measure how often you make mistakes, but how much each mistake costs.

4. Ease of implementation: can it be automated now?

This is the criterion most inflated by optimism. A task is easy to automate when the process is always the same, the data is already digital and the tool you use can connect to others. It's hard when every case is different, the data lives on paper or in someone's head, or the whole team has to change the way it works.

Photo by Mikhail Nilov on Pexels

The scale: what each score means

So that two people in the same business score things the same way, it helps to pin down what each number means. This is the scale I use:

Note that time is scored per repetition, not per month. That way you don't count frequency twice. Hours per month are calculated separately, as a common-sense check.

Before you start: take an inventory of repetitive tasks

Before scoring you need a list. The most reliable way to build it is to note down everything that repeats during a normal week. Not a holiday week, not Christmas week: an ordinary week.

1. Write it down as you work

   Keep a notebook or a note on your phone at hand and log each repetitive task as you do it. If you leave it until the end of the day, you'll forget half of them.

2. Describe each task with a verb

   "Confirm tomorrow's appointments", "send invoices to the accountant", "answer price questions on WhatsApp". One task per line. If a task has several parts, split them: they usually score differently.

3. Time it at least three times

   Don't guess how long it takes. Measure three repetitions and use the average. Most people underestimate their short tasks and overestimate their long ones.

4. Ask the rest of the team

   If several people work in the business, each has their own invisible list. Whoever is at the front desk knows things the owner never sees.

Ten or fifteen tasks on the list are enough to get started. It doesn't need to be complete; it needs to be real.

The template: a table to fill in with your own data

Copy this table into a spreadsheet or print it out. Fill in one row per task.

The columns are calculated like this:

• Hours per month = times per month × minutes each ÷ 60. This is your sanity check: if a task comes out with high priority but takes half an hour a month, you've scored something wrong.

• Priority = (Frequency + Time + Errors) × Ease. The result ranges from 3 to 75.

A filled-in example: a hypothetical car repair garage

To show you what it looks like, here's an invented example —not a real client— of a three-person car repair garage in rural Galicia. The figures are reasonable for a garage that size, but they're illustrative: yours will be different.

There are several interesting things in this table.

First, what the owner wanted to automate doesn't come out on top. Answering WhatsApp is the task with the most hours (five a month), but each answer is short, a slip is rarely serious and, to automate it properly, you first have to write the official answers to the frequent questions. It lands in the second round. That doesn't mean a WhatsApp agent is a bad idea; it means that, in this garage, two other things pay off sooner. If you want to understand what an agent can and can't do, we cover it in what a WhatsApp agent does (and doesn't do).

Second, appointment reminders win clearly. They happen every day, a missed reminder leaves an empty slot nobody pays for, and the rule is simple: "one day before, send a message with the time". It's the kind of automation that's quick to set up and noticeable from the first week. If you run an appointment-based business, this article on no-shows explains how to measure what missed appointments cost you.

Third, there's a tie between "car ready" and invoices, both at 36 points. When there's a tie, my rule is to start with the lower-risk task. Telling someone their car is ready is almost impossible to break; preparing invoices for the accountant involves tax data, and a badly built automation there produces exactly the kind of error you want to avoid. Better to get to it with some experience and a properly reviewed process, especially with Spain's VERI*FACTU changes on the horizon.

Fourth, ordering parts comes last despite adding up to more than five hours a month. Each supplier has its own website, phone line or sales rep, and there's no single way to connect to all of them. It can be improved, but it's not the first job.

The top three tasks add up to around twelve hours a month. For a three-person garage, that's almost a day and a half of work recovered every month, without touching anything complicated yet.

Photo by Mikael Blomkvist on Pexels

How to read the result

Once the table is filled in, sort the rows from highest to lowest priority and split them into three groups:

There's one extra rule worth respecting: if a task has an ease score of 1 or 2, it doesn't go first even if its total is high. The groundwork comes first: getting the data into digital form, deciding who does what or choosing a tool. That work is valuable too, but call it what it is and budget for it accordingly.

Common traps when scoring

After running this exercise with several businesses, the same mistakes keep showing up:

• Guessing minutes instead of measuring them. "That takes two minutes" is almost always five once you count finding the data and getting back to what you were doing.

• Scoring ease optimistically. If automating something first requires three people to agree and a database to be migrated, it's not a 4.

• Forgetting the exceptions. A task that's the same 80% of the time and needs human judgement the other 20% can be half-automated. Score the repetitive part and leave the exception to a person.

• Automating a broken process. If nobody knows exactly how a task is done today, automating it will only make the mess move faster. Define the process first, then automate it.

• Ignoring maintenance. Every automation needs reviewing: a price, opening hours or a message template changes. Plan for it from day one.

How much it makes sense to invest in each automation

The table tells you what to automate first. To know how much to spend, there's a quick calculation: hours saved per month × what your hour is worth × 12 months. That's the value of one year of that task running on its own.

For example, if appointment reminders take you four hours a month and you value your time at 25 euros an hour, that's 100 euros a month, around 1,200 euros a year. Use your own figure: what you charge per hour, or what it would cost to pay someone to do it. If a solution costs clearly less than that in the first year, maintenance included, it's reasonable. If it costs more, or someone tells you it "pays for itself in three years", be wary: in a small business too much changes in three years to rely on that horizon.

The idea isn't new. Cartoonist Randall Munroe summed it up in a well-known xkcd strip, "Is It Worth the Time?", which crosses how often you do a task with how much time you'd shave off each time. It's a handy way to eyeball whether an automation is worth it.

And remember that not every benefit is measured in hours. A task with a high error score can justify the investment even if it saves little time, because the real saving lies in the mistakes you stop making. That's the case with quote follow-up: every forgotten quote is a job someone else probably takes.

Photo by Kampus Production on Pexels

What to do after filling in the table

With your ranked list in hand, the path looks like this:

1. Pick a single task

   The highest one on the list with an ease score of 3 or more. Just one. The temptation to start three at once is strong, and it's the fastest way to finish none.

2. Write the process in five lines

   What triggers the task, what data it needs, what gets done, who gets notified and what happens if something fails. If it doesn't fit in five lines, it's probably two tasks.

3. Measure before changing anything

   Write down how much time you spend today and how many errors there are. Without that baseline, a month from now you won't know whether the automation helped.

4. Roll it out and review after four weeks

   Let it run for a month and compare with what you measured. Adjust whatever needs it: messages, timing, exceptions.

5. Go back to the table

   Update the scores. Sometimes automating one task makes another easier because the data is already in order. Pick the next one.

In short

Deciding what to automate first doesn't take intuition or expensive tools. It takes a week of note-taking, half an hour of scoring and the honesty to accept that the task that annoys you most isn't always the one you most need off your plate.

Frequency, time, errors and ease. Four numbers per task, one multiplication and a ranked list. With that, the first automation picks itself, and it will most likely be less spectacular than you imagined and a lot more useful.

If you fill in the table and want a second opinion on the order, or you're not sure how to score the ease of a task, get in touch and we'll go through it with you.