Short answer: Automate work that happens often, follows the same steps every time and moves information between tools, such as copying enquiries into a CRM or building the Monday report. Keep a person in the loop where money leaves the business, something goes to a customer that can't be taken back, or a decision affects a person.
When a team is stretched, the obvious answer is to hire. Sometimes that's the right answer. But before you write the job ad, it's worth looking at what the new person would actually spend their days doing. If a good part of it is copying data from one tool into another, sending the same email with a different name in it, or building the same report every Monday, you may be about to hire someone to do a machine's job.
That's bad for you, because it costs more than it should. It's also bad for them, because nobody took a job hoping to retype invoice numbers.
What work is worth automating?
Good candidates usually share a few traits. The more of these a task has, the better.
- It happens often. Daily or weekly, not twice a year.
- It follows the same steps every time. If you can write the steps on a sticky note, a machine can probably follow them.
- It moves information between tools. From a form to a spreadsheet, from the CRM to the invoicing system, from an email to a task list.
- Mistakes come from boredom, not judgement. A typo in an amount, a field left empty, a customer emailed twice.
- People complain about it. The tasks your team quietly hates are usually the ones worth looking at first.
Some everyday examples
- A new enquiry comes in through the website and someone copies it into the CRM, then emails the right salesperson. That can happen on its own.
- An order is paid and someone creates the invoice, updates the stock sheet and sends a confirmation. Same.
- Every Monday someone exports numbers from three tools and pastes them into a report. The report can build itself and land in your inbox.
- A new employee joins and someone creates their accounts in six places, from memory, slightly differently each time. A checklist that runs itself is more reliable than a memory.
- Customers ask the same five questions by email. Drafting the reply can be automated, even if a person still reads it before it goes.
How to find them in your own business
Ask each person on the team to keep a rough note for a week of anything they do more than once that feels mechanical. Not a time-tracking exercise, just a list. At the end of the week you'll have a pile of candidates, and the same few will show up on several people's lists.
Then for each one, ask three questions:
- How often does this happen?
- How long does it take each time?
- What happens when it goes wrong?
Start with the task that happens most often and takes longest. The third question tells you how careful to be with it.
When should a person stay in the loop?
Automation is very good at doing the same thing correctly a thousand times. It's also very good at doing the wrong thing a thousand times, quickly, before anyone notices. The better question is what happens when it's wrong.
Keep a person in the loop wherever:
- Money leaves the business. Payments, refunds and supplier invoices above a level you're comfortable with. Let the machine prepare them, and let a person approve them.
- Something goes to a customer that can't be taken back. A contract, a price, a reply to a complaint. A draft is fine. Sending it is a human's call.
- The decision affects a person. Hiring, firing, credit, anything with legal weight. Under GDPR, people in the EU also have rights around decisions about them made purely by automated means (Regulation (EU) 2016/679, Article 22), so it is also a legal point.
- The input is messy. Free-text emails, scanned documents, anything where the format changes. AI tools can help a lot here, reading and sorting, but they also get things wrong with great confidence. Have them suggest, and have a person confirm.
- The rules change often. If the process changes every month, the automation will need changing every month too, and someone needs to own that.
A useful pattern is "machine prepares, human approves". The tedious part disappears, and the decision stays with someone who can be asked about it later.
Things that go wrong with automation
- Nobody knows it exists. It was built by someone who has since left, and when it breaks, nobody knows where to look. Write down what each automation does and who looks after it.
- It fails silently. A tool changes its login or its data format and the workflow quietly stops. Every automation should tell someone when it fails.
- It automates a bad process. If the process itself is confused, automating it just makes the confusion faster. Sometimes the best fix is removing a step, not automating it.
- It's on someone's personal account. Build it on accounts the business owns, so it doesn't vanish when one person does.
So, automate or hire?
Often both. Automation takes away the repetitive part of a job, and what's left is the work that actually needs a person: talking to customers, solving the odd problem, making judgement calls. You might find you still need to hire, but for a different role than you expected. Or you might find the team you already have can breathe again.
Either way, it's worth knowing before you sign an employment contract.
Sources
- Regulation (EU) 2016/679 (GDPR) on EUR-Lex, Article 22, on decisions based solely on automated processing.
- Data Protection Commission (Ireland), Your rights in relation to automated decision making, including profiling, a plain-language explanation of the same right.
If there's a task your team copies between the same two tools every week, our automation page explains how we approach it. We connect the tools you already have, through their APIs or a tool like Make or n8n, and we start with whatever eats the most hours. Or book a free call and walk us through it.