What to automate in a small business, and what not

Many small businesses know they lose time on repetitive things, but keep putting automation off. Either it seems too big a project, or they are afraid something will break. Below we show you how to choose what is worth automating and where to start. And if you want us to look at your processes together, our 30-minute call is free.

What does it mean to automate a process?

It means a step that a person used to do by hand now happens on its own. The step can be very small, for example a few fields copied from one file into another, or it can be a whole workflow.

For a long time, automation only worked on fixed rules: a certain type of data went in, a certain result came out. Real-life data, though, does not always look the same. One customer writes the address in a single field, another puts the street and the number separately, a third forgets the number. For a fixed rule, every new variation was a problem.

AI models changed that. They understand what the data means, not only what it looks like. They work out that "12 Republicii Street" and two separate columns, "Street" and "Number", say the same thing, and they put the information into the form you need. This means many tasks that until recently could only be done by hand can now be automated.

How can you tell if a task is worth automating?

It is worth automating if it repeats and takes your time. If you do it daily or every week, it is a candidate. If it eats whole hours, it becomes a priority.

The maths is simple: put side by side how long the task takes you and how long it would take to automate. Say a task takes you an hour and you do it once every two or three weeks. If automating it takes three or four hours, you have recovered the time in two or three months. After that, the hour stays yours, and you can put it into something that actually grows the business.

Before you decide, watch the task a few times. Write down every step, from first to last, and who does it. Often it is only on paper that you see where the time goes.

A good sign: if you can explain a task step by step when someone asks what you do, it means you do it often. You already know how it goes, and that makes its automation much quicker to prepare.

Then put the task through three questions:

1. Does it repeat? Daily, weekly, at the start or at the end of the month. If yes, go on.
2. Is the time saved bigger than what the automation costs? Work out how long it takes you per month and how soon you recover what you invest.
3. If it goes wrong, is it easy to fix? If a mistake means lost money, a wrong contract or something that cannot be undone, a person checks the result before it goes any further.

What does a small business usually automate?

Most often: sorting messages, copying data, documents made from the same template and checking texts. The table below shows what is worth it and why.

This is what we do at Let's Automate: together we find the tasks that take most of your time and we automate them, piece by piece.

A real result

A company we work with received PDFs from its customers, each in a different format, and copied the values by hand into its internal Excel files. Transcribing one document took around an hour. Now the data fills in on its own, and the team checks it in five minutes.

What is better left to people?

Contact with the customer, money, contracts and any decision that is hard to fix. The customer wants to talk to a person, and a mistake in a contract is not solved with an undo button.

Behind the person, AI can be very useful. It has access to the company's documentation and can give hints to whoever answers customers: where the answer is, what was done in a similar case, what the next step is. From what we see, an AI model gets things wrong far less often when it is given the company's information than when it answers from what it knows in general. The face of the company, though, remains a person.

The same goes for texts. AI can check whether an announcement says what it should and whether it is written correctly. Don't let it become the company's voice. Its role is to check that the text sounds like you, not to write in your place. We wrote separately about what AI does well and where a person is needed.

A complicated process does not necessarily stay with people. If it repeats often, that is exactly where you save the most time, as long as you automate it in pieces. What usually stays with people are only the complicated processes that come up rarely, where the preparation would take longer than the time saved. What is rare but simple is often worth it, as you saw in the calculation above.

One automation, or everything connected together?

Start with a single piece. A whole process is automated in one go only if it is very small and one person does it.

When a process goes through several apps, uses several types of data or takes a long time, split it up. First automate the road from A to B, then from B to C. This way:

• You see exactly where you gain. You know which piece cut the time and the effort, and by how much.
• You try without risking much. You find out whether it is worth it before investing a lot of money or a lot of time. Every owner has different priorities, so the gain feels different from one business to another.
• The team keeps up. Most people are not technical and need time to get used to what changes. One piece is learned much faster than a whole workflow.
• The team comes with ideas. Whoever uses the first piece sees best where more can be improved.

The road to a fully automated process is a little longer. In exchange, you see the first result much sooner.

What can go wrong, and how do you protect yourself?

The two most common problems are that the team does not use the automation and that situations come up that nobody foresaw. Both shrink a lot if you automate in pieces.

• The team does not use it. Many businesses have a fear of automation: what if something breaks? And when an automation is not used, most of the time it is because people did not understand it. It is much easier to understand 10% of a new process than 100% at once.
• Unforeseen situations. In programming they are called "edge cases": the cases at the margin, which come up rarely and nobody thought of. The more you automate at once, the more of them there are. And when something breaks in a big workflow, you don't even know where to start looking.

No automation is perfect from day one. What you can control is how big the piece is in which you look for the problem. So, slowly and surely: one piece, then checking it in real work, then the next one.

Frequently asked questions

Does automation mean I let people go?

No. Their role changes: instead of copying data, they oversee the process and have the big picture. Automation gives you room to work better, do more and sell more, and the growth does not come from people being let go.

How long does it take to set up an automation?

A small, well-defined piece can be ready in a few hours. A workflow started from scratch usually takes 1 to 2 weeks, and most of the time goes into understanding the process and into testing.

Do I need a new program, or does it connect to what I already have?

Most of the time, the automation connects to the programs you already use. When data is scattered in many places, a central place can make sense: a small app made for you, where the data comes together and from where the automations start.

---

Want to find out what is worth automating in your business? Book a free call. In 30 minutes we look at your processes together and you leave with a plan and three quick wins.