You would think an automation, at worst, is simply useless. The reality is less kind: a badly framed automation costs more than no automation at all. It ties up money upfront, it demands upkeep, and above all it makes you trust a process that, one day, gets it wrong without warning. After dozens of projects, it is always the same five mistakes that come back. None is technical: they are mistakes of judgement, and you can see them coming. Knowing them is already avoiding them.
Mistake 1: automating an unstable task
It is the most common and the most expensive. An automation is built on a stable rule: when this happens, do that. If the rule changes every quarter — because your business shifts, because regulation evolves, because you are still finding your way of working — the automation is obsolete before it pays for itself. You pay to build it, then you pay to change it at every turn, and the maintenance cost eats the gain. The question to ask is not "is it tedious?" but "will this task be done the same way a year from now?". If the answer is no, wait for the process to settle. Automating shifting ground is pouring concrete over sand. A freelancer still testing prices every couple of months should not yet automate their quotes: each rate change would force them to rebuild the setup.
Mistake 2: forgetting the cost of upkeep
An automation is not a purchase, it is a rental. The day it is built, it has a price; every month after, it has a cost. The services it connects change their terms, the tools update, your business evolves and what fit in January no longer fits in September. Whoever counts only the upfront price is setting up a nasty surprise. The right sum always adds maintenance: an automation that saves you two hours a month but demands one to stay alive only saves you one. That is no reason to give up — it is a reason to know before you sign. It is the difference between buying a tool and adopting a pet: the expensive part is not the first day, it is all the others.
Mistake 3: automating what you never wrote down clearly
You cannot automate a process you cannot explain. It is obvious, and yet it is a classic trap: someone asks to automate a task that, in reality, lives in one person's head, with its exceptions, its shortcuts and its never-formalised "it depends". When you try to put it into a machine, you discover it never had a clear rule — and the machine demands a clear rule. The useful work therefore begins before the technical part: writing the task as it is really done, exceptions included. Often, that exercise alone reveals that half the process was pointless, and that the other half simplifies. Writing the process down is tedious and unglamorous, but it is precisely the step that separates a solid automation from a house of cards.
Mistake 4: aiming at everything, rather than a piece
The urge is understandable: since you are automating, you might as well automate it all at once. It is almost always a mistake. A big project that touches everything is slow to build, hard to test, and when something goes wrong you don't know where. A small, well-chosen piece, by contrast, is up and running in a few hours, checked immediately, and delivers a measurable service from the first week. The right approach is the opposite of intuition: start with the most tedious and most isolated piece, install it, let it run, and add the next only once the first is proven. You move faster in pieces than in a block. Nobody builds an entire house before checking the foundations hold; with automation, the same caution applies.
Mistake 5: not measuring before, so judging nothing after
If you don't know what the task cost you before, you will never know whether the automation was worth it. It is the quietest mistake, because it breaks nothing: it simply leaves you without judgement. Without a starting baseline — the real time spent, the number of errors, the average delay — you are left with an impression, and impressions are wrong. Before touching anything, write down the number that matters for that specific task. It is a minute's work that later turns a hunch into proof — and tells you, in black and white, whether to continue, correct or turn back. Measuring first is not bureaucracy: it is the only way that, six months from now, the answer to "was it worth it?" is a number and not an argument.
A real case: all five in one project
The five rarely come alone. A case we see often brings them all together. A business decides to automate its invoicing just as it is switching accounting software (mistake 1: unstable base). Nobody writes down how invoicing really works, with its eyeballed discounts and its special cases (mistake 3: unwritten process). The request is to automate it all at once — invoices, collections, reminders, bookkeeping — in a single big project (mistake 4: aiming at everything). No one notes how long it took before, so nobody will know if it improved (mistake 5: no measurement). And the quote mentions only the build, never the upkeep (mistake 2). Six months later, the software has changed, half the automation no longer fits, and the general feeling is that "automation doesn't work". It is not that it doesn't work: it is that it was done with all five traps set. Each one alone is easy to avoid; together, they sink the project.
The common thread of the five
None of these mistakes is a tool mistake. They are all preparation mistakes: automating too soon, without writing the rule, without measuring, without counting upkeep, aiming too wide. The technical side almost always sorts itself out. It is the judgement beforehand that makes the difference between an automation that pays off for years and one that costs money in silence. The good news: that judgement does not require being an expert. It requires slowing down for five minutes and asking the right questions before spending the first euro.
How to know which ones affect you most
Not every business trips over all five equally. If your activity changes a lot from year to year, trap 1 is your biggest risk and it is better to automate only the most stable. If you work from memory, with no written processes, trap 3 is yours, and the first step is not technical but putting down in writing how you do things. If you are one of those who want to solve it all at once, watch trap 4 and force yourself to start with a single piece. And if you have never measured what anything costs you, traps 2 and 5 stalk you at once: you will lack the figure both to decide and to judge. Knowing your weak point is worth more than knowing all five by heart, because it tells you where to put your attention before spending.
How to avoid all five at once
There is a simple move that guards against all five at once: pricing the task before automating it. Pricing forces you to look at stability (mistake 1), to include upkeep (mistake 2), to describe the process so you can estimate it (mistake 3), to isolate a specific piece (mistake 4) and to set a starting baseline (mistake 5). It is no coincidence: putting numbers on a task forces you to answer every question that makes a project fail. That is exactly why we always start there, before a single technical line.
What to do if you have already made one
If you recognise yourself in one of the five with an automation already in place, don't delete it on impulse. Measure first: is it saving you something, even less than promised, or is it costing more than it returns? With that figure, the decision is simple. If it still helps, keep it and note what needs fine-tuning. If it costs more than it gives, switch it off without drama: an automation that doesn't pay for itself is not a personal failure, it is information. What matters is not dragging for years a process you already know isn't worth it, merely because it cost something to build. The money spent doesn't come back by keeping it alive. Re-measuring what you have automated every so often is the best way to keep the five traps from staying with you forever.
Most failed automations fail not through lack of technical skill, but through excess of haste. The five mistakes are alike: they happen when you skip the thinking step to jump straight to the tool. We have prepared a free diagnostic that puts numbers on a task in two minutes — and, in doing so, puts you through the five-mistake test without even thinking about it. Four questions, result on screen, no commitment. If it concludes the task is not ready, it will tell you — and you will have avoided the costliest mistake of all: automating something that should not yet have been automated.