How to tell when a process is worth automating

The most common wrong decision in automation isn't a technical one, it's a matter of selection. What gets automated is the process that hurts the loudest — the one the team complains about most. That is almost never the same process that costs the most. A task that causes two stressful hours once a quarter sticks in the memory. A task that eats up twelve minutes every morning vanishes into the routine — and, over a year, adds up to many times as much.

So before talking about tools or models, one sober question is worth asking: how do you even recognise that a process is a good candidate? From nearly a decade of automation in Fortune 500 environments, and from what we at Skopa first automated in our own operations, five reliable signs emerge.

1. It runs often, and always the same way

The strongest lever is frequency. A process that runs daily, or several times a week, quickly pays back the time spent building it. A process that occurs two or three times a year almost never does — there, a good checklist is worth more than a system. Just as important: the flow must be largely the same each time. Rule-based, recurring steps can be modelled cleanly; a process that looks different on every run isn't a process at all, but a series of individual decisions.

2. The rules can be put into words

If the person carrying out a task can explain in a few sentences the rules by which they decide, the process is a good candidate. If the honest answer is "I just know it when I see it," there's experiential knowledge behind it that's hard to capture in rules. That doesn't make it impossible — but it shifts the effort from the automation itself to understanding the process. That is precisely why we start by observing, not by building.

3. The data is already structured

Automation is only as good as the data it rests on. A process whose inputs come in a clearly structured form — in a system, a database, a clean export — can be automated reliably. If the data is scattered across email attachments, handwritten notes and three unreconciled spreadsheets, the first step isn't automation but tidying up the data foundation. That is often the more honest and more rewarding place to begin.

4. Mistakes are expensive or embarrassing

Manual work is tiring. In recurring reconciliations, transfers and checks, errors creep in that can't be blamed on anyone — they are the price of monotony. Where a single overlooked mistake becomes expensive or costs customers' trust, automation pays off twice over: it saves time and lowers the error rate at the same time. A system compares the two-hundredth invoice with the same care as the first.

5. It blocks or delays something

Some processes cost less through the work itself than through the waiting they create. An approval step that three other people are waiting on. A report that only appears at month-end, even though the decision would need it weekly. When a manual step becomes a bottleneck, the real gain from automation isn't the hour saved but the decoupling: things happen when they're needed, not when someone finally finds the time.

When it's better not to

Honesty is part of the analysis. There are cases where automation is the wrong answer — and selling it anyway would be unserious. In those cases we say no, and we write down the reasoning, so the question doesn't come back around in a year's time.

• The process runs rarely. What happens twice a year rarely justifies the build time — a documented checklist is cheaper and more robust.
• The process is about to change. Automate a flow you'll overhaul in six months anyway, and you build it twice. Sort out the process first, automate afterwards.
• The exception is the rule. When almost every run is a special case, you end up automating the handling of exceptions above all — expensive and error-prone.
• The human contact is the value. Some flows are deliberately personal. Making them more efficient would mean rationalising away the very thing that makes them valuable.
• The data foundation is missing. Without reliable, structured data, you automate the chaos — faster, but no better.

«Doing nothing isn't free. But automating everything isn't either.»

The most expensive process is the invisible one

One pattern recurs in almost every analysis: the processes with the greatest leverage are rarely the ones on a wish list. Whatever someone flags as "annoying" is already on their radar — if pushed, they could organise it differently. The expensive ones are the flows that have become so self-evident that no one questions them any more: the daily copying together of figures, the weekly reconciliation of two systems, the manual re-entry because two programs don't talk to each other. This work doesn't feel like a problem, it feels like "the job." And that is precisely where its treachery lies.

That is why, in an analysis, we deliberately speak with the people who carry out the process day to day, not only with management. The truth about a workflow is rarely known by the person who described it, but by the one who does it every day. Only when you see the process where it actually happens does it become clear where the time is really lost — and not merely where its absence is felt loudest.

From a hunch to a number

The five signs are a compass, not a proof. Whether a specific process pays off is decided by the numbers: how often per month, how long per run, how error-prone, who waits on whom. Only from that does a solid figure emerge — hours per year, valued at your internal costs, not at an industry average pulled from a study.

We walked this path first in our own business — not by automating everything, but by choosing the right things. That choice is exactly what sits at the heart of every potential analysis: an honest ranking of what pays off first, what can wait, and what you're better off doing by hand.