Business Automation: What to Automate, and What to Leave Alone

The short answer

Business automation means handing a repetitive, rule-based task to software instead of doing it by hand each time. It only pays off once the process underneath is sound, because automation repeats a process exactly, flaws included. The real question isn't whether a task can be automated. It's whether it should be, and what that frees your team to do instead.

23 Sept 20267 min read

What automation takes off your plate

It's the end of the day, and you're doing the part of the job nobody sees: copying an order from an email into a spreadsheet, typing the same details into the accounting system, sending the "just following up" email you've sent a hundred times before. None of it is hard. It just never stops.

That's the kind of work automation handles well. Anything that follows the same rule every time can usually be handed to software:

The task	What automation does	What sets it off
Re-typing details between systems	Copies the information across once it's entered	A new order, booking or form
Standard emails and reminders	Sends the right message at the right time	A date, a status change, a missed deadline
Routine documents	Builds the invoice or report from data you already have	A finished job, the end of the week
Chasing approvals	Routes the request to the right person and follows up if it stalls	A request over a set amount

Set up well, that can mean fewer evenings spent on admin, fewer typos creeping in between systems, and less depending on one person remembering to do everything. If you're trying to work out which of your own repetitive tasks is actually worth tackling first, how to identify automation opportunities walks through ranking them by real cost rather than gut feel.

The aim isn't to replace anyone. It's to take the repetitive work off your team so their time goes to what needs a person: decisions, customers, and the problems no rule can solve.

Why "can we automate this?" is the wrong first question

Almost anything repetitive can be automated, so "can we automate this?" nearly always gets a yes. The more useful question is whether the process, as it runs today, is worth automating, and what actually makes it suitable for automation in the first place.

Automation copies a process exactly, including the parts that shouldn't be there. Cut these three before you automate anything:

A report nobody reads. Before automating a weekly report, ask who opens it and what they decide from it.
An approval that's never refused. If the answer is always yes, a clear rule, such as "only amounts over a set limit need sign-off", may do the same job.
A double-check for a problem that's gone. Find out why the step was added. If the original problem no longer exists, neither does the reason for the check.

One caution: some checks are controls, not habits. A second sign-off on payments, for example, often exists so no single person handles a payment from start to finish. Those stay, whether or not the steps around them are automated.

If you haven't mapped and redesigned the process yet, start there. This guide picks up once a process is in reasonable shape: what to automate, what it costs, and what to leave alone.

Rules or AI: which kind of automation does a task need?

Most business automation is rule-based: when this happens, do that. Rules are predictable and easy to check, which makes them the right starting point for most small businesses.

AI is useful for a narrower job: the step where the input is messy and someone currently has to read it and decide what it means. For example:

Reading a supplier invoice PDF and pulling out the amount, due date and ABN
Sorting incoming emails into enquiries, orders and complaints
Drafting a first reply for a person to check before it's sent

Two practical rules keep AI useful rather than risky:

Start with fixed rules. Add AI only for the step that genuinely needs interpretation.
Treat AI as an assistant to your team, not a replacement for it. It can read, sort and draft so nobody starts from scratch, but a person stays responsible for the result, especially wherever a mistake would cost money or affect a customer.

Where the usual answer falls short

Lean has a word for what automation shouldn't touch

Lean manufacturing calls a step that adds no value muda. Toyota's approach is often summed up as improving the manual process before automating it, because a machine repeats a wasteful step as faithfully as a useful one. A small business doesn't need the vocabulary to use the idea: redesign the process first, then decide what's left to automate.

Start here: four steps before you automate anything

1
Start with the process, not the tool
Map how the process runs today and cut anything that shouldn't be there. Automating a process nobody has looked at locks in whatever's wrong with it.
2
Pick one repetitive, low-judgment task
Something done the same way, often, with little real decision-making involved. Not "our whole invoicing process", but one specific step inside it.
3
Work out what it costs today
Minutes each time, times how often it happens in a year, times the hourly cost of whoever does it. The yearly figure is usually bigger than it feels day to day.
4
Decide: automate, outsource, hire, or leave it
Automation is one option among several, not the default. The check further down this page covers the signs a process isn't ready yet.

Is your process ready to automate?

Run a process through this quick check before you spend time or money automating it. A process is usually ready when:

Everyone describes it the same way. Two people walking you through it would list the same steps in the same order.
Most cases follow the standard path. Exceptions happen, but they're the minority, and you know what they are.
You know roughly what it costs today. Even a rough yearly figure gives you something real to compare against.
The decisions in it can be written down as rules. "If the invoice is under $500 and it's an existing client, approve it" is a rule. "It depends on the client" isn't.

If one of these isn't true yet, that's not a no. It tells you what to sort out first:

People describe it differently? Map it against a real, recent example.
Exceptions everywhere? Redesign it so the standard path covers most cases.
No cost figure? Work out the cost and payback before comparing options.
A genuine judgment call in the middle? Keep a person on that step, and look at automating the steps around it instead.

Example: what this looks like for invoicing

Here's how the four steps above apply to a typical invoicing process. No invented business and no made-up numbers this time, plug in your own at each step.

Map it. A typical run looks something like: job marked complete → job sheet reaches the office → invoice typed into the accounting system → approved → invoice emailed → payments checked → reminder sent if overdue. Map your own version against a real, recent invoice; the order and the handoffs matter more than the exact steps above.
Remove before automating. Check whoever approves invoices: do they actually refuse any, or does every one get signed off regardless of amount? If it's always a yes, replace the approval with a rule instead of automating a rubber stamp, for instance, sign-off required only above a set amount, with everything below it going straight through.
Cost it. Time how long creating and sending one invoice actually takes, and how many reminders go out in a typical month and how long each one takes. Add those two together, multiply by 12 for a year, divide by 60 for hours, then multiply by what an hour of that person's time genuinely costs your business. That gives you a real annual figure, not a borrowed one.
Decide what to automate. In most invoicing processes, creating the invoice once a job is marked complete, emailing it, and sending reminders on a fixed schedule are rule-based and good automation candidates. Disputes and payment-plan requests usually need a person.

Whatever your own number comes out to, that's what you weigh against the cost of automating. The automation ROI guide shows how to work out the payback using your real figures.

From the checklist to your specific process

Once a process passes that check, the question changes from "should we automate this?" to "which steps, exactly, and in what order?" That depends on your systems, your volume, and which parts of the process are genuinely rule-based rather than judgment calls that only look like rules on paper, and it's easier to work out on a real process than a general one.

Walk away knowing exactly what to automate first

Bring one process, invoicing, onboarding, whatever's costing you time, to a free discovery call with our Principal Consultant. You'll leave knowing which of its steps are worth automating and which aren't, and if it's a fit, what it would look like running in your business's own M-Street System workspace.

Book a free discovery call

Explore Business Automation

You don't need to read these in order. Jump to whichever matches the decision in front of you.

What Is Business Process Automation?What automation means, how it differs from workflow automation, and what makes a process suitable for it.How to Identify Automation OpportunitiesA practical way to find and rank the tasks in your business worth automating.What to Automate First, and What to Leave AloneHow to choose between automating, outsourcing, hiring, or leaving a task exactly as it is.How to Calculate Automation ROIA simple way to work out whether an automation project pays for itself, and how fast.When Does a Spreadsheet Become a Business System?Spreadsheets aren't the problem on their own. What matters is whether the data in them is reliable enough to automate anything on top of.Where AI Actually Helps, and Where It Doesn'tThree practical questions to decide whether a specific step needs AI, a plain rule, or a person.Automating a Process Yourself, or Having It Built For You?An honest comparison of the real costs and risks of each approach, and which one actually fits your situation.What Happens When an Automation Fails?Why automations actually fail, the difference between a loud failure and a silent one, and what's worth checking before trusting one with something that matters.Why Are We Entering the Same Data Twice?Why the same information ends up typed into more than one system, and how to find and fix it.

Frequently asked questions

What's the difference between business automation and workflow automation?

They're often used interchangeably, but business automation is the broader term for using software to handle any task without manual effort. Workflow automation specifically means automating the flow of a multi-step process, connecting individual automated tasks so work moves between them without someone handing it on by hand.

Is my business too small to automate anything?

Probably not. Size isn't the deciding factor; repetition is. A one-person business that sends the same follow-up email fifty times a month has a genuine automation candidate. A fifty-person business whose work is mostly one-off and judgment-heavy might have very little worth automating yet.

Should I automate a process before or after I've mapped it?

After. Map and redesign the process first, then automate what's left. Automating an unmapped process usually makes an unexamined set of steps run faster, not better.

What kinds of tasks are good candidates for automation?

Tasks that are repetitive, rule-based, and frequent: moving data between systems, sending standard emails and reminders, creating routine invoices or reports, and routing approvals against a clear threshold. Tasks that need judgment, negotiation, or a personal relationship are usually poor candidates.

Can AI automate my business processes?

Some steps, not whole processes. AI is useful where someone currently has to read and interpret something messy, like an email, a PDF invoice or a free-text form. Most routine automation doesn't need AI at all, because fixed rules handle it more predictably. Think of AI as an assistant to your team rather than a replacement for it: a person stays responsible for the result, especially where a mistake would cost money or affect a customer.

Do I need technical skills to automate part of my business?

Not always. Simple automations, like a reminder or a form that feeds a spreadsheet, are often within reach of anyone comfortable with everyday software. Anything that connects several systems, or has to handle exceptions reliably, usually takes more setup and testing, and that's where having it set up for you tends to save time.