Automation · AI · Small business
What you can automate in a small business
Re-typing data, the same emails over and over, invoices by hand. Here is which tasks genuinely repay automating — and which are better left to a person.
Automation has a bad name in small businesses, because it sounds like corporate rollouts costing hundreds of thousands. In reality, the biggest time sinks are mundane: copying the same thing from one place to another, sending the same message again and again, filling in documents by hand.
This piece is about which of those genuinely belong to a machine — and, just as importantly, which do not.
How to spot a good candidate
A task worth automating meets three conditions at once:
- You do it regularly. Once a year is not worth it. Every day very much is.
- It is the same every time. If each case needs a judgement call, that is not a task to automate — it is a task to simplify.
- A mistake is catchable. An automated process will occasionally do something stupid. You need to notice before the customer does.
If any of those is missing, it is usually cheaper to leave it to a person.
Five things that pay for themselves fastest
Copying data between programs. An order arrives by email, you copy it into a spreadsheet, then into invoicing software. The same thing three times, three chances to get it wrong. This is the textbook case where automation pays back in weeks.
The same messages to customers. Confirming a job, reminding about an appointment, asking for a review afterwards. You write them from scratch every time, even though only the name and date change.
Documents. Invoices, handover notes, contracts. If they are built from data you already hold somewhere, filling them in by hand is doing the work twice.
Reports. How much did I do this month, what is unpaid, which jobs are still open. If you sit down once a month and total it up from scraps of paper, that is exactly the work a computer does without getting tired.
Product descriptions and photos. This is where AI earns its place. If you sell dozens of different things, writing a description for each one kills your pace. The same goes for product photography — I showed this with Fotarobota, where a phone snapshot becomes a studio shot without a photo session.
What not to automate
Conversations where money is at stake. Quoting an unusual job, handling a complaint, negotiating. Customers sense an automated reply, and it puts them off at precisely the moment you can least afford it.
Things you do rarely. Automating a quarterly task costs more than it saves, and it will break between one use and the next.
A process you do not understand yourself. If you cannot write the steps on paper, an automated system will not guess them. Tidy up first, automate second — otherwise you set the mess in concrete.
The last contact before a buying decision. Where a customer is choosing between you and a competitor, a person beats a machine.
How long it takes, and where to begin
Not with the biggest problem. With the most repetitive one.
Pick a single task you do every day that looks the same each time. It can usually be handed over in a few days, and you will feel the difference immediately. Only once that one piece works is it worth thinking about the next.
The opposite order — a "let's automate the whole company" project — usually ends with nothing working after three months and everybody discouraged.
How I know this works
I built a system that carries stock from a shipping manifest to a live listing: pricing it, writing the descriptions, cataloguing it. What used to eat entire evenings now happens in the background. I wrote it up in detail in the Charon case study, including the parts that turned out harder than I expected.
The same thinking transfers to a small business in any trade. The point is not to replace the person — it is to stop them re-typing.
If you want to test this on your own business
The simplest experiment: for one week, note down everything you do on the computer more than once a day. The list that comes out of it is a better starting point than any guide.
Then, if you like, describe it in the order builder. A few plain questions, no jargon — I will tell you honestly what can be taken off your plate and what is not worth it. If it turns out the real problem is the website itself rather than your processes, I will say that too.