Getting started
What should a small business automate first?
Almost everyone picks wrong on the first attempt, and they pick wrong in the same way: they automate the task they hate. The task you hate is usually rare, full of judgement, and expensive to get wrong — which is close to the worst possible candidate. The right first automation is almost always something you find too boring to have considered.
The answer, before the reasoning
Automate the job you repeat most often, that has the same shape every time, and where an error costs you a minute rather than a customer.
That is the whole rule. Everything below is how to apply it to your own list, and why each part of it matters more than it sounds.
Why hating a task is the wrong signal
The jobs that generate the most irritation tend to share three properties. They come up unpredictably, so there is no rhythm to build on. They vary every time, so there is no fixed shape to encode. And they usually matter — a quote, a complaint, a proposal — so they carry real consequences when they go wrong.
Frustration measures how unpleasant something is, not how automatable. Those are unrelated. The end-of-quarter task you dread is fourteen decisions in a trench coat, and every one of those decisions is you.
Meanwhile the thing you do eleven times a day without noticing — reformatting an order into the shape your supplier wants, pulling four numbers into the same weekly summary, writing the same three-line reply with two details changed — is invisible precisely because it is regular. Regular is the property that makes automation work.
Score your own list in ten minutes
Write down every repeated job you can think of. Not projects — jobs. Then score each one from 1 to 5 on three axes.
Frequency: how often does it happen?
Daily is a 5. Weekly is a 4. Monthly is a 2. Anything less often than monthly is a 1, and honestly should come off the list. Something you do four times a year will not repay the setup, and by the fourth time you will have forgotten how it works.
Sameness: how identical is it each time?
Ask whether you could write down the steps and hand them to a new starter with no explaining. If yes, that is a 5. If the honest answer is it depends, work out what it depends on. Two or three known variations is still a 4 — you can encode branches. "It depends on the customer" is a 1, and that is the real answer to why it has not been automated already.
Cost of a mistake: what happens when it gets it wrong?
This one is scored backwards, and it is the axis people skip. A 5 means an error is trivial and obvious — a badly formatted internal summary, a draft you were going to read anyway. A 1 means an error reaches a customer, moves money, or quietly corrupts a record nobody looks at until it matters.
Score the mistake honestly, not optimistically. "It will not get it wrong" is not a score. Every automation gets something wrong eventually; the question is only what happens on the day it does.
Multiply the three numbers. Start with the highest total. In almost every business the winner is something mundane and slightly embarrassing to admit you were doing by hand.
A worked example
Here is the same list from a small services business, scored properly.
| Job | Frequency | Sameness | Cheap to get wrong | Total |
|---|---|---|---|---|
| Turning an enquiry email into a CRM entry | 5 | 5 | 5 | 125 |
| Drafting the reply to a standard enquiry | 5 | 4 | 4 | 80 |
| Weekly figures pulled into one summary | 4 | 5 | 4 | 80 |
| Chasing overdue invoices | 3 | 4 | 3 | 36 |
| Writing a custom quote | 3 | 1 | 1 | 3 |
| Year-end accounts preparation | 1 | 2 | 1 | 2 |
The quote and the year-end are the two jobs that owner most resents. They score at the bottom, and correctly so — both are mostly judgement, and both are expensive when wrong. The winner is a task nobody complains about because it only takes ninety seconds. Ninety seconds, eleven times a day, five days a week, is more than an hour every week going into retyping.
Three things not to automate first
- Anything that sends without you seeing it. Your first automation should produce a draft you approve, not an action it takes. The confidence to remove yourself is earned after a few weeks of watching, not assumed on day one.
- Anything that moves money. Payments, refunds, invoices going out. The error mode is not embarrassment, it is a bank transaction, and you will be arguing about it for a fortnight.
- Anything where a mistake is invisible. A wrong entry written into a database nobody reads until quarter end is worse than a loud failure, because by the time you find it there are ninety of them.
Measure it before you build it
Spend one week counting. How many times did that job happen, and roughly how long did it take? You want two numbers before you start, because otherwise you will have no idea whether the thing you built was worth building.
This sounds like bureaucracy and takes about four minutes a day. It is the difference between "that felt useful" and "that removed seventy minutes a week, and here is the note I wrote down before I started".
It also protects you from the more common outcome, which is an automation that saves six minutes a week and costs twenty minutes a month to maintain. That happens more often than anyone admits, and without the before-figure you will never notice.
How long the first one should take
If your first automation is going to take more than a day to build, you have chosen something too ambitious. Pick a smaller piece of it.
The purpose of the first build is not the time it saves. It is that you learn what breaks, what your data actually looks like when it arrives, and how much checking you need before you trust it. That knowledge is what makes the second and third ones fast, and it is only available by finishing one.
A finished small thing beats an unfinished ambitious thing by a margin that is difficult to overstate. The graveyard of small business automation is full of half-built systems that were going to handle everything.
One line to keep
The first job to automate is not the one that makes you sigh. It is the one so routine you stopped noticing you were doing it — and it is probably on your list already, sitting near the bottom, scored too low because it never seemed worth mentioning.