What one automated process actually saves
Most conversations about automating something never get as far as a number. They stay at the level of “that would save so much time”, which is true, useless, and the reason half these projects never start and the other half get approved without anyone checking whether they were worth it.
The sum is not complicated. It just has four terms instead of the two people usually consider, and the two extra ones are where the answer actually lives.
The sum
How often does it happen. Per week, not per year. Something weekly beats something monthly by roughly four to one and the gap compounds over the life of the automation.
How long does it take. All of it. Including finding the login, waiting for the export, and the ten minutes afterwards spent getting back into whatever you were doing before. People underestimate this by about half, consistently, and the last part is the bit they leave out entirely. Task switching is not free and it belongs in the number.
What is that hour worth. Not the person’s salary divided by 1800. The loaded cost, which for most European small companies lands somewhere well above the gross wage once you include everything the company actually pays. Your accountant can tell you this figure in about four minutes and it is worth knowing for reasons beyond this exercise.
What does it cost to keep running. This is the term everyone forgets. Automations are not free after they are built. There is a platform subscription, there is model usage if AI is involved, and there is the occasional afternoon when an upstream system changes and something breaks. Call it a small monthly figure plus a few hours a year.
Multiply the first three. Subtract the fourth. That is your annual return, against a one time build cost.
A worked example
Let me be explicit that the numbers below are arithmetic, not a case study. Put your own in.
Say a weekly report takes ninety minutes, including the fifteen minutes of scattered attention afterwards. That is 78 hours a year. At a loaded cost of 25 euros an hour, that is just under 2000 euros a year of someone’s time.
Say the automation costs 30 euros a month to run, which is generous for something this size, and needs four hours of attention a year. Round the running cost to 500 euros.
Net saving is roughly 1500 euros a year. Against a build that costs, say, 1000 euros once.
It pays for itself in about eight months and then keeps going. Not transformational. Real.
Where this sum surprises people
Upward, on frequency. Owners consistently underrate daily tasks and overrate annual ones. A fifteen minute daily task is 65 hours a year. The quarterly reporting ordeal everyone complains about for a fortnight beforehand is maybe 40 hours, and it is the one that gets the attention because it is painful rather than because it is large.
Downward, on one off work. If it happens twice a year, it almost never clears the bar no matter how much everyone hates it. Automating it will cost more than doing it will, forever. The correct answer is to do it badly by hand and stop thinking about it.
Sideways, on who is doing it. The same task has a different answer depending on whose hours it is eating. Forty hours a year of administrative time is a minor saving. Forty hours a year of the owner’s time, in a company where the owner is the bottleneck on everything, is worth considerably more than the arithmetic suggests, because those hours are not fungible.
What the sum leaves out
Two things, both real, neither of which I would put a number on.
Automated work does not drift. A person doing the same task 50 times will do it 50 slightly different ways, and some of those ways contain a mistake nobody catches until a customer does. Consistency has value that does not show up in hours saved.
And things that run by themselves run on time. Reports that arrive Monday morning rather than when someone remembers change how decisions get made, in a way that is genuinely hard to quantify and that people notice within a month.
I mention these because they are real and I refuse to put a euro figure on them, which is a distinction worth holding onto when someone else does put a figure on them.
The point of doing this
Not to justify the spend. To pick the right thing.
You will end up with several candidates. The sum is what tells you which one to do first, and it very often disagrees with the room. The process everyone complains about is rarely at the top. The thing at the top is usually something small, frequent, and so unremarkable that nobody had thought to mention it.
Do that one, live with it for a month, then do the next.
If you want to run this over your own company with someone who has done it before, book a call. Thirty minutes, and you keep the numbers either way.