Automating a broken process gives you a broken process that runs faster.
That one sentence would save Australian small businesses a lot of money if more owners heard it before signing up for software. The process improvement vs automation question comes up constantly in our operations work, and the honest answer is that most processes need fixing first, some need automating after, and a few need to be deleted entirely. Here’s how to work out which one you’re looking at before you spend anything.
Fix first, automate second, delete where you can
The rule we work to is simple. If a process produces errors, rework or arguments about whose job it is, fix the process. If it produces the right result but eats hours of repetitive human effort, automate it. And if nobody can explain why the process exists, kill it and see what breaks.
Automation is a multiplier. It multiplies whatever you feed it, good or bad. Feed it a clean, well-defined process and you get back hours every week. Feed it a messy one and you get the same mess, produced at scale, with a monthly subscription attached.
We see this pattern often in reviews. An owner buys workflow software because the admin is chaotic, then six months later the admin is still chaotic and there’s a new tool nobody fully uses. The software wasn’t the problem. The process underneath it was never defined.
How do you know a process is broken rather than just manual?
A broken process fails even when people follow it correctly. A manual process works fine, it just takes human time to run. That distinction decides which way you spend your money.
Signs a process is broken:
- Two people run it and get different results
- The output regularly needs rework or checking by someone senior
- Steps exist that nobody can explain, they’re just “how we’ve always done it”
- It depends on one person’s memory rather than anything written down
- Handovers between people are where things go missing
Signs a process is merely manual:
- The result is right almost every time
- Anyone trained can run it and get the same output
- The steps are the same every time, no judgement calls in the middle
- The main cost is hours, not errors
Be honest about which list your process sits in. Owners tend to describe broken processes as “just slow” because slow feels like a software problem and broken feels like a management problem. It’s usually the second one.
Four questions to ask before you spend a dollar
1. Can you write the process down in ten steps or fewer?
If you can’t, you don’t have a process, you have a habit. Automation vendors will happily quote you for automating a habit, and the project will stall the moment their tool hits an undocumented exception. Map it first. A whiteboard and half an hour with the person who actually does the work is enough.
2. Is the output right at least nine times out of ten?
Anything below that and you’re looking at a quality problem, not a speed problem. Fix the failure points before you touch automation. Often the fix is boring: a checklist, a single template instead of five, or moving one step earlier in the sequence so information arrives before it’s needed rather than after.
3. How many hours a week does it actually consume?
Measure it, don’t guess. Owners routinely overestimate the pain of annoying tasks and underestimate the pain of frequent ones. A monthly report that takes three frustrating hours costs you 36 hours a year. Rekeying job details that takes ten minutes, twenty times a week, costs you roughly 160 hours a year. The second one is duller and far more expensive.
4. Would the automation pay for itself inside twelve months?
Do the maths on paper. If setup plus a year of subscriptions costs more than a year of the labour it replaces, wait, or fix the manual version instead. Twelve months is a fair test for a small business because tools change, staff change, and long payback periods rarely survive contact with reality.
A worked example with real numbers
Take a Brisbane trades business where the office manager rekeys job details from customer emails into the job management system.
- 20 jobs a week, about 10 minutes of rekeying each
- 20 × 10 = 200 minutes, call it 3.5 hours a week
- At $50 an hour fully loaded, that’s $175 a week
- Over 48 working weeks, roughly $8,400 a year
An automation quote might land around $5,000 to build plus $150 a month to run, so about $6,800 in year one. Payback inside twelve months, and the process is consistent and error-free. That passes the test.
Now the fix option. Replace free-text emails with a structured booking form, so job details arrive in the right fields to begin with. That might be a day of setup and no ongoing cost, and it removes most of the rekeying anyway. Cheaper, faster, and it improves the customer’s experience too.
In this case the fix wins, and the automation conversation can wait until volume grows. That’s the point of doing the sums first. The right answer changes with the numbers, not with what a software demo makes you feel.
Fix or automate at a glance
| What you’re seeing | Likely answer | First move |
|---|---|---|
| Inconsistent output between staff | Fix | Write the process down |
| Frequent rework or corrections | Fix | Find the failure step |
| Right result, heavy repetitive hours | Automate | Cost the hours, get quotes |
| Nobody knows why it exists | Delete | Stop it, watch for a month |
| High volume and already consistent | Automate | Test payback under 12 months |
Mistakes we keep seeing owners make
Automating around one person. If the process lives in one employee’s head, the automation gets built to match their quirks, and when they leave nobody can maintain it. Document first, then automate the documented version.
Buying the tool before defining the job. Software demos are persuasive. A demo shows you a clean process running in a clean system, which is not your business on a Tuesday afternoon in October. Decide what the process should do, then find a tool that does that, in that order.
Ignoring the exceptions. The happy path might be 80 per cent of cases. The other 20 per cent, the change requests, the partial payments, the customer who rings instead of emailing, is where automations quietly fail. Ask any vendor how their tool handles your exceptions specifically, and watch how quickly the answer gets vague.
Counting saved hours nobody redeploys. Automation only pays off if the freed time goes somewhere useful, whether that’s billable work, sales calls or simply not hiring the next admin person as early. If the saved hours just evaporate, so does the return.
Common questions about process improvement and automation
What are the 5 key elements of process improvement?
Define the outcome you want, map how the work actually happens now, measure time and error rates, remove the steps that add no value, then standardise what’s left so everyone runs it the same way. Most small businesses skip the mapping step and pay for it later.
What are the 5 stages of BPM?
Business process management typically runs through design, modelling, execution, monitoring and refinement. For a small business, the practical version is simpler: write the process down, run it consistently, measure it, and improve it when the numbers tell you to.
What are the 7 steps of BPR?
Business process re-engineering is the heavy-duty version, where you redesign a process from scratch rather than improving the existing one. The formal frameworks run through steps like setting objectives, mapping the current state, identifying gaps, redesigning, and implementing. Most businesses under $10 million in revenue don’t need formal BPR. They need their five worst processes written down and cleaned up.
Which automation tool is the best?
There’s no single best tool, and tool choice matters far less than how well the process underneath is defined. Many businesses can automate their first few workflows using features already inside software they pay for, such as their accounting platform or job management system, before buying anything new.
Should I automate a process before improving it?
No. Automating a flawed process locks the flaws in and makes them more expensive to change later, because now the fix involves rebuilding the automation as well. Improve first, run the improved version manually for a few weeks, then automate what’s proven.
Is it cheaper to fix a process or automate it?
Fixing is almost always cheaper upfront, often costing nothing beyond a few hours of thinking and a rewritten checklist. Automation costs more initially but wins on high-volume, repetitive work where the payback period is under a year.
How do I measure whether an automation paid off?
Record the hours the manual process consumed before you build anything, then compare against total automation cost after twelve months, including setup, subscriptions and the time spent maintaining it. If nobody measured the before state, you’ll never know, which is exactly how underperforming software survives in businesses for years.
Work out where your hours are actually going
If you’ve got a process that’s eating time and you’re not sure whether to fix it, automate it or bin it, that’s a conversation we have most weeks. Our operations work starts by mapping what’s really happening in your business, putting numbers against it, and telling you honestly which problems are worth spending on. Sometimes the answer is software. Often it’s a better checklist. Get in touch and we’ll look at yours together.