What to automate first in a small business

Most small businesses do not need artificial intelligence or a new software platform. They need the same three or four jobs to stop eating hours every week. This guide covers how to spot those jobs, the ones that are usually worth automating first, how to work out whether it pays, and what to leave alone.

Start with the task, not the tool

The most common automation mistake is starting with a product. Someone hears about a tool, signs up, and then looks for something to use it on. Six months later the business is paying for another subscription and still doing the work by hand.

Start the other way round. Write down the jobs that come back every week and annoy you, your staff or your customers. Do not worry about how they could be automated yet. Just get the list.

A job is a good candidate when it has three things in common:

• It repeats. Daily or weekly, not once a quarter.
• It follows the same steps each time. If you could write the steps on a card and hand them to a new staff member, a computer can probably follow them too.
• It involves copying, chasing or checking. Typing the same details into two systems. Sending the same reminder. Checking whether something arrived.

If a job ticks all three, it belongs near the top of the list.

The usual first candidates

Every business is different, but the same jobs come up again and again.

Enquiries from the website. A form arrives in someone's inbox, gets copied into a spreadsheet or CRM, and someone replies to say thanks. All three steps can happen on their own: the enquiry goes straight into the system you use, the customer gets an immediate reply that tells them what happens next, and the right person is notified.

Quote follow-ups. Quotes go out and nobody follows them up, because nobody has time. A follow-up that sends itself a few days after the quote, and stops when the customer replies or accepts, wins work that was otherwise quietly lost.

Invoice reminders. Chasing overdue invoices is unpleasant and easy to put off. Accounting software such as Xero can send reminders on a schedule. If yours is not set up, that is often the cheapest automation you will ever do.

Booking confirmations and reminders. A confirmation when the booking is made and a reminder the day before cut down on no-shows and on phone calls asking "what time was I booked in?"

Re-entering data between systems. Orders typed from the website into the accounting system. Job details copied from email into the job management tool. Customer details entered twice because two systems do not talk to each other. This is where hours disappear without anyone noticing, and where mistakes creep in.

Weekly reporting. If someone spends Monday morning pulling numbers from three places into one spreadsheet, the report can be built for them and waiting when they arrive.

How to tell whether it will pay for itself

You do not need a business case. You need a rough sum.

1. Estimate the time. How many hours a week does the job take, across everyone who does it?
2. Put a cost on that time. Use what that person's hour costs the business, or what you would charge a customer for it.
3. Multiply by a year. Hours a week, times the hourly cost, times the weeks you work.
4. Compare it with the cost to automate. Include the build and any ongoing subscription.

As an example: if a job takes three hours a week, and an hour of that person's time is worth $50 to the business, that is $150 a week. Over 48 working weeks, it is $7,200 a year. An automation that costs a fraction of that, and keeps running, pays for itself within the year.

Then add the things that are harder to put a number on. Fewer mistakes. Customers who get an answer straight away instead of tomorrow. A task that no longer falls apart when one person is on holiday. These are often worth more than the hours.

If the sum comes out small, that is a useful answer too. Not every annoying job is worth automating.

Off-the-shelf tools or something built for you

There are two broad ways to automate a task.

Connect the tools you already use. Services such as Zapier and Make link popular software together: when something happens in one, do something in another. For simple, common jobs they are quick to set up and good value. Many of the systems you already pay for also have automation features you are not using. Check those first.

Build something for your process. When the job crosses several systems, follows rules specific to your business, or involves software that does not connect easily, a custom-built workflow is often more reliable. It does exactly what your process needs and nothing else.

Both have costs. Off-the-shelf tools usually charge monthly, and the price can climb as you use them more. A custom build costs more upfront but can sometimes replace a subscription entirely. The right answer depends on the job, which is why it pays to describe the job first.

What not to automate

Some things are better left to people:

• Jobs that change every time. If each instance needs judgement, automation will either get it wrong or need so many exceptions it is not worth it.
• Conversations that win or keep customers. An automatic thank-you is fine. An automatic reply to an unhappy customer is not.
• A process that does not work yet. Automating a messy process gives you a faster mess. Fix the steps first, then automate them.
• Jobs that happen a few times a year. The time saved rarely covers the effort, and nobody remembers how the automation works when it finally runs.

A simple way to begin

Pick the one job on your list that is most frequent and most annoying. Write down every step, including who does it, which systems they touch and where things go wrong. Note how long it takes in a normal week.

That one page is most of what anyone needs to tell you whether the job is worth automating, and how. Often the first answer is cheaper than people expect: a setting in software you already own, or one connection between two systems.

If you would like help working it out

We build workflow automation for Perth businesses, starting with the systems you already use. For anything beyond a simple fix, we start with discovery: 3 hours mapping your process, your tools and what to automate first, for $450, ending in a written scope and a price. You can read more on our automation page, or describe the job in plain English through our contact page.