First: automating is not buying software
It's worth clarifying before the list, because almost every project that fails skips this distinction. Automating a process means getting a sequence of steps to happen without someone having to push it along: without copying data from one system to another, without chasing an approval over WhatsApp, without building the same report every Monday.
Sometimes that is achieved with a new tool. Other times it is achieved by connecting two systems the company already pays for. And with uncomfortable frequency it is achieved by removing the step, because it only existed to correct an error that another step no longer makes. The tool is the last decision, not the first.
1. There are people whose job is moving data between systems
It's the clearest sign and the easiest to quantify. Someone exports from one system, adjusts in Excel and uploads to another. Sometimes it's half an hour a day; sometimes it's a full-time person. If that person is also the only one who knows how to do it, the company doesn't have a process: it has a dependency.
To put a number on it, a single question to whoever does it is enough: how many times a week, and how long does it take each time. Multiply by four and you have the hours per month. It's a figure that tends to surprise on the high side, especially once you add up three or four similar tasks spread across different departments.
2. Nobody can say where a request stands without asking
If finding out the status of an order, a purchase order or a customer request means calling someone, the process has no traceability. The cost isn't just the call: it's that problems are discovered when it's already too late, because nobody was watching the point where it got stuck.
This sign has a telltale symptom: follow-up meetings whose only purpose is for each department to say where things stand. That meeting is a report the system should be producing on its own.
3. Approvals live in email or chat
A discount approved over WhatsApp leaves no usable trail. When someone asks six months later who authorized that price, the answer is a search through a chat, assuming the person is still with the company.
Here the cost has two sides: the time spent chasing the sign-off —which is usually the real bottleneck of the process, not the work itself— and the risk of not being able to prove the authorization when it matters. In regulated sectors, the second side is the expensive one.
4. The same data is typed in more than once
The customer's details are written on the quote, written again on the order and written a third time on the invoice. Every entry is a chance for error, and the error is discovered at the end: in the wrong shipment or the rejected invoice.
A practical rule: count how many times the same piece of data is typed by hand across the whole journey. If it appears more than once, rework is guaranteed and you're probably already paying for it in credit notes.
5. Management reports are built by hand
If consolidating the monthly report takes someone two days of copying cells, there are two problems, not one. The first is obvious: those days. The second is worse: because it's so costly to produce, nobody asks for the data between closings, and decisions are made on information that is three weeks old.
The sign that confirms this point is familiar in any committee: two departments present different figures for the same month and the meeting is spent arguing about which one is right. That isn't fixed with a new dashboard; it's fixed by agreeing on definitions and tracing where each number comes from.
6. Errors are caught late, and always from the same side
If the customer is the one who tells you the order went wrong, control is on the outside. Manual processes fail silently: there's no alert, no exception, just a case that kept moving forward with the wrong data until it ran into someone.
It's worth listing the last five errors that cost money and asking at which step they originated and at which step they were discovered. The distance between those two points is what automation shortens.
7. Knowledge of the process lives in a single person
There's someone who knows how it's really done, which exceptions exist and who to call when something goes off-script. When that person goes on vacation, the process slows down; when they resign, it becomes chaotic.
Automating well forces you to write those rules down, and that exercise has value even if nothing ends up being automated. A process that only exists in someone's head can't be improved, audited or delegated.
How to put a figure on it before asking for budget
None of the seven signs justifies an investment on its own. What justifies it is the sum, and the sum is calculated with a simple table any department head can fill in over an afternoon: for each manual task identified, how many times a month it happens, how long it takes and what an hour of the person doing it is worth.
Add to that figure what rework costs: credit notes, repeated shipments, overtime at closing. And add compliance risk where it applies, which in regulated sectors often outweighs everything else combined.
With that number on the table, the tone of the conversation changes. It stops being “we should modernize” and becomes “this costs this much a month and solving it costs this much”. The first version is a battle of opinions; the second is a decision.
- Monthly frequency of each manual task identified
- Average duration and hourly cost of whoever does it
- Rework: how many cases a month and what each one costs
- Exposure from non-compliance, when the process is regulated
Where to start without committing half the budget
Start with a single flow, and not the most complicated one. The goal of the first one isn't maximum savings: it's proving to the team that this works and learning how your operation reacts to change. A repetitive process with stable rules and high volume is the ideal candidate.
Before automating it, redesign it. The question isn't how to do what's done today faster, but which steps would no longer be needed if the process were designed from scratch. Automating a poorly designed process only makes errors happen faster and at higher volume.
And define in advance how you'll know it worked. One metric, measured before starting. Without a baseline, at the end of the project nobody will be able to prove what improved, and the next investment will once again be a battle of opinions.