Get the order right and everything after it gets easier. The first job wants to be one that comes round often, cannot do much damage if it slips, and is visible to the people doing it — because what you are really building first is confidence.
Pick the first one well and the second and third take half the effort, with nobody needing to be talked into them. That is what the first one is really for.
The instinct is to start with the biggest job — the one that eats the most time, or the one that would look most impressive if it worked. It is a perfectly fair instinct. It also makes the first attempt harder than it needs to be, and the first attempt is the one that quietly decides whether there is ever a second.
A better first job passes three tests.
Imagine a job that comes round every Tuesday morning. Within a fortnight you have watched it run twice, found the two things nobody thought of, and fixed them. Within two months it is simply how Tuesday works now.
Now imagine the same effort spent on something that happens once a quarter. Same build, same cost, and you will not know whether it was worth doing until next year. The rough edges — and there are always rough edges — take three months each to show up.
Frequency is also what makes the arithmetic work. Four minutes, twenty times a week, is more than a full week of somebody’s year. It never feels like it while it is happening, which is exactly why jobs like that sit there unnoticed for years.
Anything new gets things wrong at the start. The question is never whether that will happen, it is what it costs when it does.
So the first job wants to be one where a wrong answer is spotted in seconds and put right in seconds. A draft that somebody reads before it goes anywhere. A summary that sits next to the original. A file that gets filed, where the wrong place would be obvious at a glance.
What it should not be is the thing that goes out of the door with nobody reading it. Not because that cannot be built safely, but because it is the wrong place to learn. Start with a person checking the output as a matter of course, and decide later, from evidence, what is safe to leave to itself. Keeping that order is more or less what doing it safely comes down to.
This is the test that gets skipped, and it is the one that decides everything.
Suppose the first thing you automate quietly saves four hours a month somewhere nobody sees. On paper it worked. In practice nothing has changed, because nobody’s week feels any different and nobody has a reason to believe the next idea will land either.
Now suppose instead that the person who used to spend Tuesday morning on a job gets Tuesday morning back, and can see exactly why. They will mention it to the person next to them. When you come back with the second idea you are not making an argument any more — you are being asked when it will be ready.
If people are wary at the start, it is usually for a sound reason: somebody has handed them new software before and it made the week harder rather than easier. Nothing you say will shift that. One small, visible win that is obviously theirs will.
The first thing you automate is not really about hours. It is about whether anybody believes the second one.
The third and fourth things. Almost always.
By then a few things have changed. You know what your own information looks like in practice — where it lives, how it is named, which parts are reliable and which are not — and that groundwork does not have to be laid twice. People trust the output enough to stop checking every single line, which is where a surprising amount of the time saving actually appears. And everybody has a much better instinct for what is worth handing over, because they have now seen one working.
So the first one does not need to be impressive. It needs to be finished, used, and obviously fine. The ones that pay for themselves several times over are the ones that come after it, and they only come if the first one landed.
There is one more argument for starting small. A first project that takes a fortnight is allowed to be wrong. A first project that takes four months is not, and the pressure to call it a success whatever the truth does nobody any favours.
Which leads to the part that usually gets left out, and it is worth saying plainly: plenty of good first jobs need nothing built at all.
A fair number of the best first wins are a saved template, a reminder set up once, or a feature switched on in software that is already on your bill. Business software has changed a great deal in the last two years and most of it now does more than the person who bought it realises. If your first win costs nothing and takes an afternoon, that is a better first win rather than a lesser one — it proves the same point at a fraction of the price. What we do starts in the same place: find it first, and build only what genuinely needs building.
If you want to do this on your own, the method is short. Write down the jobs that come round every week. Cross off any where a mistake would be quiet and expensive. Of what is left, pick the one that the most people can see. Start there, finish it properly, and let everybody watch it work for a month before you choose the next one.
Then pick the next one, and the one after that. Those are the ones you are doing all this for.
Two pieces that pair well with this: a way to find where your own hours actually go before you pick, and why the tool matters less than the job you point it at.
Which job is the right first one depends entirely on how your own week runs — working that out, in order, with a number against each, is most of what the audit is for.
Twenty minutes on the phone, free, no obligation. Tell us the jobs that come round every week and we will tell you which one we would start with, and why.