Every week, a growing business does the same small tasks hundreds of times. Typing an order into a sheet. Sending an invoice. Reminding a customer to pay. Checking whether stock is running low. Copying a number from one place to another. Each one takes a minute or two. Together, they can take up a meaningful part of someone's week, and they're where a lot of errors creep in.
Automation means getting software to do some of these tasks without a person each time. Done well, it frees people for work that needs judgment and reduces mistakes. Done badly, it produces confusion faster than any person could. The difference lies mostly in what you automate, and in what order.
What I want to do here is give you a simple way to decide what to automate, in what order, and what to keep in human hands.
The rule that matters most
Automate a process only once it's clear, consistent and working.
The reason is simple. Automation repeats whatever you tell it to, exactly, every time, at speed. If the process is clear, that's a gift. If the process is muddled, with steps people handle differently, exceptions nobody has written down, information that's sometimes in one place and sometimes another, automation repeats the muddle, and makes it harder to see and fix, because now it's happening inside a system instead of in front of you.
So the sequence is: understand the process, write it down, make it consistent, measure it, and then automate the parts that are purely mechanical. (Our guide to standard operating procedures covers the first steps.)
Does that make sense? Think of a photocopier. Give it a clean original and it produces a thousand clean copies. Give it a smudged original and it produces a thousand smudges, very quickly. Automation is a photocopier for processes.
What makes a task a good candidate
The best tasks to automate share a few features.
They happen often. A task done fifty times a week is worth automating; one done twice a year usually isn't.
They follow the same steps every time. No judgment needed, or only very simple rules.
They involve moving or copying information. Re-typing data from one place to another is slow, boring and a common source of errors.
Mistakes in them are costly. A missed invoice, a forgotten reminder, a stock-out on a key line.
Where small businesses often start
A few areas tend to repay automation early.
Invoicing. Generate invoices directly from confirmed orders, with the right prices, customer details and due dates, so nothing waits for someone to type it up and nothing gets missed.
Payment reminders. Send a polite reminder automatically a few days before an invoice is due, and another on the due date. This is one of the cheapest ways to get paid faster, because it makes the routine consistent. (See our guide to getting paid on time.)
Order confirmations and updates. An automatic message when an order is received, when it's dispatched, when it's delivered. Customers ask fewer "where's my order?" questions, and your team answers fewer of them.
Stock alerts. A warning when a key line falls to its reorder point, so ordering happens on time instead of when someone notices the shelf is empty.
Bank and payment matching. Many accounting tools can import bank transactions and suggest which invoice each payment belongs to, which makes reconciling the books much faster.
Reports. The monthly snapshot of sales, margins (the share of each sale left after paying for the goods), debtors and cash, generated from your records instead of assembled by hand.
WhatsApp quick replies and away messages. Not automation in the full sense, but they take seconds to set up and make sure every enquiry gets an immediate, consistent first response.
Keep a person where judgment lives
Some things shouldn't be automated, or only partly.
Anything that needs judgment: approving a large discount, deciding whether to extend credit to a new customer, handling a complaint.
Anything where the cost of a wrong automatic action is high: payments to suppliers, for example, should be prepared automatically if you like, and still approved by a person.
Anything where the personal touch is the point. A long-standing customer who's been with you for years may notice the difference between an automatic message and a call from you, and you may want them to.
A good pattern is to automate the preparation and keep a person on the decision. The system drafts the invoice, prepares the payment run, flags the overdue account; a person checks and approves.
There's a chain behind a failed automation project. The disappointment is usually the result of a tool that doesn't fit how the business works. The poor fit is the result of choosing the tool before understanding the process. Start from the process instead, and the chain runs the other way.
Choosing tools
There's a lot of software available, and it's easy to buy more than you need. A few principles help.
Start from the problem, then look for the tool. "We spend six hours a week typing orders into the accounts" is a problem a tool can solve. "We should get a CRM" is a tool looking for a problem.
Prefer fewer tools that connect to each other over many that don't. Every extra system is another place information has to be entered and kept up to date.
Check that your data stays yours. You should be able to export your customer, sales and financial records if you ever change tools.
Pilot before committing. Run a new tool on one process, for a month, before rolling it out everywhere.
Count the full cost. Subscription fees, setup time, training, and the time it takes to keep it maintained.
Measure the effect
Before you automate a task, note how long it takes and how often it goes wrong. A month after, check again. That tells you whether the automation is paying for itself, and gives you the case for the next one. It also catches the occasional automation that creates more work than it saves.
The bigger point
Owners tend to move through three views of automation. First, excitement: automate everything, buy the tools. Then, after a few tools nobody uses, scepticism: technology is a waste of money. Then a steadier view: automation amplifies whatever system you give it. A clear, well-run process becomes faster and more reliable. A muddled one becomes faster and more muddled. So the real work is in the process, and automation is the last step that multiplies it.
How Gridwerk helps
Mapping your processes, deciding what's ready to automate, choosing tools and setting them up so the automation fits the way your business works is part of our Process Optimization service. If your team spends too much of its week on repetitive tasks, start with the free diagnostic. It's eighteen questions, about seven minutes, and then we call you to go through the rest.
Gridwerk diagnostic