Which task do you automate first? A calculation for your working week
Not the task you hate most, but the one that comes back often, goes the same way every time and does little damage if it goes wrong once. Here is how to work out where to start.
New to this topic? Start with Automation for solo businesses: 8 tasks that can run themselves from tomorrow
The wrong first choice
People who start automating usually pick the task they dislike most. The bookkeeping, the quote for that one difficult customer, the monthly report you always dread.
That is understandable and usually a poor first choice. The tasks you dislike most are often also the tasks with the most exceptions. Slightly different every time, with decisions you only make once you see the situation. Those are exactly the hardest to automate well.
The best first task is more boring. It comes back often, goes roughly the same way every time, and if it goes wrong once, it is quickly fixed.
Four questions per task
For one week, write down which recurring tasks you do. Then ask four questions per task:
| Question | Why it matters |
|---|---|
| How often does it come back? | Daily delivers more than once a quarter |
| How much time does it take per run? | Ten minutes a day is well over forty hours a year |
| Does it go the same way every time? | The fewer exceptions, the more reliable the automation |
| What happens if it goes wrong? | An error in an internal overview is less serious than a wrong email to a customer |
The first two tell you what it delivers. The last two tell you how safe it is to start.
The calculation
Time per year is easy to work out: minutes per run, times the number of runs per week, times the weeks you work.
| Task | Per run | Per week | Per year (46 weeks) |
|---|---|---|---|
| Going through the inbox and pulling out important mail | 15 min | 5 times | 57.5 hours |
| Preparing tomorrow's appointments | 10 min | 5 times | 38 hours |
| Making a weekly overview of numbers | 45 min | 1 time | 34.5 hours |
| Review request to last week's customers | 20 min | 1 time | 15 hours |
| Monthly report for a client | 2 hours | once a month | 22 hours |
Those numbers are an example; fill in your own times. What stands out: the small daily tasks add up to more than the big monthly job you dread.
Score on predictability and risk
Next to the time per year, add two simple scores, from low to high:
- Predictability. Does the task follow the same pattern every time? A daily overview of your inbox does. Negotiating a contract does not.
- Risk of an error. Does an error stay internal, or does it go out? An overview for yourself is low risk. An invoice to a customer is high.
The best first candidate has a lot of time per year, high predictability and low risk. In the example above that is the daily inbox overview or the preparation of your appointments. Not the monthly report, and certainly not the invoices.
Start with reading, not with sending
There is a second rule that prevents a lot of disappointment: first automate what reads and summarises something, and only then what sends or changes something.
An automation that summarises your inbox and calendar every morning can break very little. If the overview is incomplete once, you notice and look yourself. An automation that sends emails on its own can send ten customers a wrong message in one go.
That is how you build trust. After a few weeks you know how reliable the reading work is, and you can take the step to something that sends more deliberately, with a check in front of it.
Examples per type of business
- Freelancer in business services: every working day at eight, an overview of unread mail and that day's appointments.
- Web shop: every Monday, an overview of last week's revenue and failed payments.
- Installer: every Monday, last week's customers on a list for a review request.
- Agency: every month, an SEO audit per client site, with the differences compared with last month.
All four are reading work or work with a check in front of it, come back often and go the same way every time.
After the first: the next one
Once your first automation has run for a month without you needing to look at it, take the next one from your list. That way, in six months you build up a handful of automations that together save hours a week, without anything big ever going wrong.
How to combine them into a way of working that runs itself is covered in automation for solo businesses.
How it works in dGENIX
dGENIX has a simple rule: anything you can ask GENI by hand, you can also schedule. You write the instruction the way you would type it in the chat, choose a frequency, and GENI carries it out at the time you set. Daily, on working days, weekly or monthly.
"Every working day at eight, summarise my unread mail and put today's appointments underneath."
Irreversible actions ask for confirmation even inside a scheduled task. So no email or publication ever goes out unseen, not even when you decide to take the step to sending.
The same goes for the work of the engines: the SEO Engine can run an audit every month, the Reputation Engine can send the review requests every Monday. More about scheduling is in scheduled tasks explained.
Frequently asked questions
How much time does a task need to take to be worth automating?
There is no lower limit, but a task that takes less than a few minutes a week delivers little. Calculate it per year: from a few hours a year it starts to be worth it.
What if a task sometimes goes the same way and sometimes does not?
Automate the part that is predictable and keep the exceptions yourself. Making an overview can be automatic; the decision on what to do with it stays with you.
Should I automate everything at once?
No. One task at a time, and only the next once the previous one has run reliably for a few weeks. That way, when something goes wrong, you immediately know where it came from.
