How to choose the first task to automate in your business
A simple test for choosing your first task to automate: how often it happens, who does it, what a mistake costs. With good and bad examples.
Pick a task that happens at least every week, that one person already does in a fairly fixed way, and where a mistake is cheap to spot and fix before it reaches a customer. Drafting replies to booking emails, or turning a pile of supplier invoices into a weekly spreadsheet, usually passes that test. Payroll, tax filings and quotes on large orders usually fail it. Below is the test we use when a company asks us where to start, with examples of first choices that tend to work and ones that tend to go wrong.
How often does the task happen?
Frequency is where the time savings come from, so do the sum before anything else. Take the minutes each round takes and multiply by how many times it happens in a week.
Say a café gets around 15 messages a day asking about opening hours, gluten-free options and group bookings, and each reply takes two minutes. That is half an hour a day, or three and a half hours a week if you open every day. Small per message, large over a year.
Compare that with the annual stocktake. It might take two full days, but it happens once a year, and by the time it comes round again your suppliers, your shelves and your spreadsheet have all changed. A task that runs daily also shows you within a week whether the automation works. A yearly one leaves you waiting twelve months to find out.
As a rough rule, aim for something that happens several times a week at the very least.
Who does it now, and can they explain it?
The best first task belongs to one person who does it the same way each time. Ask them to talk you through the last five times they did it. If they can describe the steps in ten minutes and the five examples look alike, you have a routine. If every case needs a different judgement call, it is still a skill living in someone's head, and that is a poor place to start.
Watch out for tasks three people do three different ways. In an architecture practice, each partner might write fee proposals in their own style with their own spreadsheet. Automating that means first getting the partners to agree on one way, which is a meeting, not a software project. Do the meeting first.
The person doing the task should also want it gone. They are the one who will check the first outputs and tell you what is wrong, and they will only do that properly if the result saves them time.
What does a mistake cost?
This is the filter that matters most for a first project. Sort the possible mistakes into three groups:
- The mistake is caught before anything leaves the building, because a person reads the draft first.
- The mistake reaches a customer but costs little, like a review reply with a slightly stiff tone.
- The mistake costs money or causes legal trouble, like a wrong VAT rate, a wrong price on an export contract or a wrong answer about allergens.
Your first task should sit in the first group. The simplest way to get it there is to set the system up so it prepares the work and a person approves it. The reply is drafted, someone reads it and presses send. The spreadsheet is filled in, someone checks the totals. Once it has been right for several weeks running, you can decide whether some of it can go out without a check.
Good first choices and bad ones
These tend to make good first tasks:
- Drafting replies to booking and enquiry emails for a restaurant, which staff read and send.
- Sorting supplier invoices and delivery notes into a spreadsheet every week, checked by whoever does the books.
- A weekly one-page report that pulls sales from the till and bookings from the reservation system, so the owner stops copying numbers across on Sunday night.
- Drafting replies to Google reviews. For restaurants and hotels, this is what Écho does inside Maison, which is free to try.
- For a shop, answering the same ten product questions that come in by WhatsApp, with a person approving each answer.
These tend to make bad first tasks:
- Payroll and tax filings. The rules depend on your country and change often, mistakes are expensive, and your accountant or gestor should own them.
- Quotes on large orders where the price depends on the customer, the season and a phone call.
- Anything that tells customers about allergens or health information without a person checking it.
- Tasks that happen a few times a year, however much you dread them.
- A process your team has not agreed on yet.
Notice that the task you hate most is not always the right one. It is often the rare, high-stakes job, which is exactly what fails the test.
Write the task down before you talk to anyone
Whether you build something yourself or bring in help, a one-page description saves a lot of back and forth. Write down what starts the task (an email arrives, it is Monday morning), what goes in, what should come out, who checks it, how often it happens, how many minutes it takes and what a mistake would look like. Attach five real examples, with names removed if they contain customer details.
With that page, you can judge for yourself whether the task is worth automating. If you want us to look at it, scope your own system asks five short questions and writes a draft proposal on the spot with our public starting price. It is a draft, not a quote: the exact scope and price come back in writing within two working days. A tailored system for one task starts from €699 one-off plus the monthly plan from €75 a month, and the price sheet has the tiers.
For this week, keep a tally sheet by the till or on your desk. Every time someone does a repetitive task, they add a mark next to it and note the minutes. On Friday, run the three questions over the top three, and you will have your first task.