In this guide
What exactly does automation change?
Every automation has three parts: a trigger, a rule and an action.
Example: an order comes in from your website (trigger). If the payment is approved (rule), an invoice is issued, stock is reduced and the customer gets a shipping message (action). Anything outside the rule, such as an order with an incomplete payment, lands on your team's desk.
So automation does not replace your staff; it stops them redoing the same task every day. People step in where the rule ends and judgement begins.
The potential is not small. According to McKinsey's 2023 research, today's technology is technically capable of automating work activities that absorb 60–70% of employees' time. That does not mean half of all work disappears tomorrow. But it is worth a look before saying "our work can't be automated".
Which tasks suit automation? A 4-question test
Put any task you have in mind through these four questions:
- Is it frequent? Does it repeat at least a few times a week?
- Is the rule clear? Could you explain the steps in writing to a new employee?
- Is the data digital? Does the information come from a form, an email, a marketplace or a spreadsheet?
- Is a mistake costly? Does getting it wrong have a price? A wrong invoice, a missed appointment, a customer nobody got back to.
If you answer "yes" to three of the four, the task is a candidate for automation.
Some tasks do not suit it: tasks that need judgement every time (price negotiations, talking to a customer in a crisis), tasks done once a month that take ten minutes, and tasks whose rule changes every week.
Tasks most often automated, by industry
| Industry | Done by hand today | Automated |
|---|---|---|
| E-commerce | Entering orders into accounting one by one | When an order arrives, invoice, stock and shipping message move on their own |
| Clinics and beauty | Phoning clients to remind them of appointments | Automatic reminder the day before; a cancelled slot opens to the waiting list |
| Real estate | Typing enquiries from listing portals into a spreadsheet | Enquiries land in one list and are matched to suitable listings automatically |
| Restaurants and cafés | Entering daily sales and stock into Excel every evening | Till data flows into a dashboard; you get an alert when an ingredient runs low |
| Manufacturing | Quote requests scattered across email and WhatsApp | Requests enter one flow; you see which quote is at which stage on one screen |
| Services and SaaS | Welcome, contract and invoice steps for a new customer | All of them start in order the moment the customer signs up |
What does it earn you? Work out your own number in 5 minutes
Calculate, don't guess. The formula is simple:
Time per task × times per month = time spent per month
Example: entering one order into accounting takes 4 minutes and 600 orders arrive a month. 4 × 600 = 2,400 minutes, or 40 hours a month. That is roughly one working week of one employee, every month.
Two tips for a more accurate result:
- Don't estimate the time; log it for a week together with the person who does the task.
- Next to the time, write down the cost of mistakes. A wrong invoice, a forgotten appointment, a customer answered too late. These often cost more than the hours.
Where to start? 5 steps
- Pick one task. Start with the one that repeats most often and that your team complains about most. Not all of them at once.
- Write down how it works today. Step by step: who does what, where the information comes from, where it goes.
- Separate the exceptions. Automation follows the rule; decide up front which cases fall outside it and leave them to the team.
- Build small and run it side by side for a week. Let the manual and automated process run in parallel and compare the results.
- Measure, then expand. Move on to the second task once the first one is settled.
Common mistakes:
- Trying to automate everything at once.
- Automating a broken process as it is. The mistakes speed up too.
- Tying the system to one person's knowledge.
- Leaving the data and access with the company that built it.
Who should keep control?
Ask these three questions before you have automation built:
- Whose server does the system run on? Yours, or the builder's?
- What happens if you part ways? Does the system keep running, or does it stop?
- Which data does it access, and who approves that?
At Epnex, the system is installed on your own server. The data, the decisions and the access rights stay with you.
When is automation not needed?
- When the task happens a few times a month.
- When the process has not settled yet: a newly founded business, or a flow that changes every week.
- When the task itself is a human relationship.
In these cases, write the process down first. Automation comes later.
Frequently asked questions
Do I need to know software to use workflow automation?
No. What you need to do is explain how the task is done today. Setup and technical maintenance are on the builder's side.
Do we have to replace our current software?
In most cases, no. Automation is usually built between the tools you already use (accounting software, e-commerce platform, Excel, email) and connects them.
Will automation take work away from our staff?
It takes away the repetitive part. Your staff spend their time on exceptions, on talking to customers and on work that needs decisions.
How long does setup take?
It depends on how many systems need connecting. A single workflow is finished much faster than a setup that links several systems. The exact time becomes clear after the analysis.
Is it worth it for a small business?
The calculation above answers that. Find out how many hours a month a task takes; if the number is large, the size of the business does not matter.
Epnex DigitalBuilds automation and AI systems that hand the work businesses do by hand every day over to software. The system runs on your own server; the data and the control stay with you.
Sources
- McKinsey Global Institute, The economic potential of generative AI: The next productivity frontier, June 2023. mckinsey.com
Last updated: 25 September 2026. Figures reflect the sources as of publication.
