What to automate first in a small business

What "automation" means here

Automation is a small program that does a job a person does now, the same way every time, without anyone pressing a button. It might run every morning, every hour, or whenever something happens, like a new order or a new form. It usually doesn't need a new app or a big system. Often it is a script that connects tools you already pay for.

Signs a task is a good first candidate

The best first automation is boring. Look for a task that is:

• Frequent. It happens daily or weekly, not twice a year.
• Rule-based. You could write the steps on one page, and two people following them would get the same result.
• Digital already. The information starts in an email, a spreadsheet, a web page, a form or another tool, not on paper or in someone's head.
• Annoying. People put it off, rush it or make small mistakes doing it.
• Low-risk if it hiccups. If it fails one morning, someone notices and nothing terrible happens.

Five common places to start

These come up in almost every small business:

1. Copying data between tools. New orders, sign-ups or bookings retyped into a spreadsheet, a CRM or an accounting tool. A script can move them as they arrive.
2. The weekly or daily report. Someone pulls numbers from two or three places and builds the same summary every Monday. A scheduled job can build it and email it to you before you're awake.
3. Watching a website for changes. Checking a supplier's prices, a listing or a public results page by hand. A scheduled check can collect it and tell you only when something changed.
4. Routine emails. Reminders, follow-ups, "your order is ready" messages. These can go out on their own, based on dates or events.
5. Cleaning up data. Fixing formats, removing duplicates and merging lists before anyone can use them.

A simple way to choose

Write down every repetitive task your team does. For each one, estimate two numbers:

• Time per month. How long it takes each time, times how often. A 15-minute task done every workday is about five hours a month.
• Cost of a mistake. Low, medium or high. A wrong number in an internal report is low. A wrong invoice to a customer is high.

Start with the task that takes the most time and has a low or medium cost of a mistake. It saves the most hours, and you get to see how automation behaves in your business before you trust it with anything sensitive. Tasks with a high cost of a mistake can come later, with a person checking the result before it goes out.

Before automating a task, ask whether it should exist at all. Sometimes the real fix is a better form, one fewer step, or a tool you already pay for that has a setting nobody turned on. That's cheaper than any script, and I'll say so if it's the case.

What to leave alone for now

• Tasks that change every time or depend on judgment, like negotiating with a supplier or handling an upset customer.
• Processes your team hasn't settled yet. Automating a process that changes every month means paying to rebuild it every month.
• Rare tasks where setup would cost more than years of doing it by hand.
• Anything involving money or customer records with no checking step. Automate the preparation and let a person approve.

What a good automation looks like

The difference between an automation that helps and one that becomes a new problem is mostly in the details nobody sees:

• It keeps a log. Every run writes down what it did, so you can check it any time.
• It tells someone when it fails. Websites change and passwords expire. A good automation sends an alert instead of failing quietly.
• Passwords are stored safely. In the operating system's keychain or encrypted secrets, never written into the code.
• Someone can understand it. A short guide on what it does and how to check on it, so it doesn't live only in one person's head.
• It runs somewhere cheap. Many automations can run on a schedule in GitHub Actions or on a computer you already have, with little or no hosting cost.

Those are the habits I build into every automation. One example: a Python bot I wrote collects results for 5 national lotteries on 4 schedules and publishes them for the Belize Lottery app. When its original data source was locked down, it was moved to a public source so it kept running. The AI tools and automation case study has more.

What it costs

A single automation is often a small project. My automation work starts at $3,000, and larger pipelines that touch several systems get a fixed quote after a free call. Projects of $3,000 or less are paid half to start and half on delivery. Changes later can go on a care plan or be billed hourly. See business process automation for what I build and how.

Getting started this week

1. Ask your team which task they'd most like to never do again. Write their answers down.
2. Estimate time per month and cost of a mistake for each.
3. Pick the top one and write its steps the way someone does it today, screen by screen.
4. Bring that list to a free call. I'll tell you whether it's worth automating and roughly what it saves.