Manual work has an awkward property: it never sends an invoice.
A software licence shows up in the budget and gets debated. A person copying data between systems for two hours a day shows up nowhere, because the salary was already there. The cost is real, but it hides inside a line nobody questions.
This article shows how to run that calculation with verifiable numbers instead of vague estimates.
What an Hour of Work Is Really Worth
The first mistake is using gross salary. What matters is the total cost to the employer, including social contributions and other charges.
According to Eurostat, average hourly labour costs in 2025 stood at €34.9 across the EU and €38.2 in the euro area, up from €33.5 and €36.8 the previous year. The spread between countries is wide: €12.0 in Bulgaria, €19.4 in Portugal, €56.8 in Luxembourg. Non-wage costs account for roughly 24.8% of the total across the EU.
So the hourly figure you should use is not salary divided by hours. It is meaningfully higher, and it varies enormously depending on where your team sits.
The Calculation Almost Nobody Runs
Take one repetitive process and apply this:
Hours per day × 220 working days × hourly labour cost = annual cost
Two hours a day of repetitive administrative work, at the euro area average, comes to roughly €16,800 a year. Per person. At the Portuguese average, around €8,500.
Across a team of five with the same pattern, that is anywhere between €42,000 and €84,000 a year depending on the country.
This is the direct cost. It is the easiest to calculate and, as it happens, the smallest of the three.
What the Benchmarks Say: The Invoice Case
Invoice processing is the most measured process in Europe, which makes it a useful thermometer.
The European Commission estimates that a broad shift from paper to electronic invoicing would generate savings in the region of €240 billion across Europe over six years. The same document points to where the lag sits: roughly 22% of SMEs exchange e-invoices, against 42% of large enterprises.
On unit cost, the studies diverge considerably. EU-funded work puts the average processing cost of a paper invoice near €30, with savings around 80% once it goes electronic. Billentis, the sector's European reference, is more conservative: roughly €11 per paper invoice issued against €4.50 electronically, with savings between 60% and 80%.
That spread is precisely why you should not trust a borrowed number.
Calculate With Your Own Figures
Use the real hourly cost from Eurostat and the time you measured. An example using Portuguese figures:
- Receiving, checking and posting a simple invoice: 8 minutes. At €19.4 an hour, that is €2.59.
- Exceptions: Billentis estimates that 20% to 30% of paper invoices end up in exception handling. Assuming 25% with 20 extra minutes, that adds €1.62 per invoice, spread across all of them.
- Realistic average: around €4.20 per invoice, in direct labour alone.
At 500 invoices a month, that is 6,000 a year and roughly €25,000 annually. Excluding systems, archiving, physical space and downstream error correction.
There is also a European figure that works as a sanity check: studies put realistic savings at 1% to 3% of turnover once invoicing becomes electronic and automated. Apply that percentage to your annual revenue and compare it with your own calculation. If both land in the same order of magnitude, the arithmetic holds.
The Costs That Never Reach Payroll
This is where the calculation above falls short.
Errors and rework. Every mistyped figure starts a chain: someone spots it, someone investigates, someone fixes it, someone confirms. Correcting always costs a multiple of getting it right first time.
Delays. An approval sitting three days in an inbox costs nothing directly. It costs when the supplier holds delivery, when the early payment discount lapses, when the customer compares your response time against a competitor's.
Opportunity cost. The largest and least discussed. Hours spent on administration are hours not spent selling, serving customers or improving the product. They do not appear as expense. They appear as revenue that never happened.
Compliance risk. Manual processes leave a thin trail. When you need to show who approved what and when, reconstructing it costs hours and sometimes proves impossible.
Turnover. Skilled people hired for skilled work who spend half their day copying data eventually leave. Replacing someone costs several months of salary in recruitment, onboarding and lost productivity.
Work Out Your Own Number in 20 Minutes
You do not need a study. You need five steps:
- Pick one process that repeats weekly. Start with the most tedious.
- Time it. Do not estimate. Ask whoever runs it to log actual time for a week.
- Multiply by annual frequency and by the real hourly cost of that role.
- Add the error rate. How often per month does something need fixing? Multiply by average correction time.
- Add waiting time. Not the work, but the days the process sits idle waiting for someone.
Steps 1 to 3 give you the number you take to the board. Steps 4 and 5 are what make the decision obvious.
For a starting point on which process to pick, we published a list of 20 processes that almost always pay off.
And What Does Automating Cost?
The comparison is only honest if it includes this side.
Software is the small slice. An n8n cloud account starts near €20 a month; self-hosted, you pay for the server, typically €5 to €10 monthly at low volumes. We covered n8n's hosting and pricing models in detail.
The real cost sits in the design: understanding the process, mapping the exceptions, building, testing and watching it through the first few weeks. That is project work, counted in days, not years.
Set that one-off investment against a recurring annual cost that repeats for as long as the process exists. That is the arithmetic that usually settles the conversation.
The Most Common Mistake in the Calculation
Plenty of companies present the saving as headcount reduction. Then nobody is let go, the promised number never shows up in the accounts, and the next project loses credibility.
The saving from an automation project is rarely money leaving the budget. It is capacity released: the same team absorbs more volume, responds faster and spends its time on work that generates revenue. We wrote about gaining that capacity without growing the team.
Frame it that way and the numbers survive scrutiny. Frame it as a payroll cut and you will be explaining, six months later, why payroll did not move.
Conclusion
Manual processes cost real money, but they cost it in a way no accounting system records. That is precisely why they survive so long.
Pick one process this week. Time it. Multiply. The number will probably surprise you, and it is close to certain that it exceeds the cost of fixing it.
The useful question is not what automation costs. It is how many years you have been paying not to do it.