What should a small business automate first?
Pick the task before you pick the tool. Five short questions help you find what is worth automating first, and what is better left to a person.
A small business has dozens of things it could automate, and most people start with the most impressive one. I suggest starting with the most boring one. The first task does not need to impress anyone. It needs to save time, and to fail small if something goes wrong.
Anthropic's guide for teams building with AI says much the same: find the simplest solution possible, and only add complexity when you need it. It adds that this can mean not building an autonomous system at all, and that fixed workflows suit well-defined tasks.
How do you choose? Five questions
This is my own method, not a standard. Run each candidate through five questions and go with the one that scores best.
1. How often does it happen? A task that comes back every day or every week can justify automation. One that happens once a year cannot.
2. How many minutes does it take each time? Multiply by how often. Ten minutes, twenty times a week, is just over three hours. That is a number you can check after two weeks.
3. How costly is a mistake? Summarizing an incoming inquiry that someone reads anyway is a cheap mistake. An invoice that goes to a customer unseen is an expensive one.
4. Is the input tidy? A form with fixed fields is easier than free-form chat messages with photos and abbreviations. The messier the input, the more checking you need.
5. Who checks the result? At the start, every result gets a human look. If nobody is going to check, it is not the first candidate.
What does a good candidate look like, and what should wait?
A good candidate, for example: inquiries that arrive by email or form, summarized into one line in a sheet, with a draft reply prepared. A person reads, corrects and sends. It repeats often, it takes time, and a small error costs no money.
Better to wait: anything that sends money or a message to a customer without a check, and any process where you do not yet have a clear definition of a good result.
And once you have picked one?
Measure before and after: how many minutes it took over the last two weeks, and how many after the automation runs. If the saving is real, expand. If not, stop and go back to the list. The articles on n8n and Make cover tools that suit a start like this.