Automation or a new hire: a cost and readiness framework for EU operations leads

How do you compare a hire with automation?

A hire is a monthly cost that starts on day one. It keeps running while the person learns your product codes and your customers' habits. Automation is a one-off setup cost followed by a smaller monthly cost. That monthly cost is never zero, because a person still checks the output.

Put both on the same chart and the question changes. It stops being "is automation good?" and becomes "in which month does the automation total fall below the hire total?" That month is your break-even point.

You need four inputs:

1. The hours the process takes each week, measured, not guessed.
2. The loaded hourly cost of the person who would do it.
3. The one-off setup cost of an automation pilot, from an itemised quote.
4. The monthly running cost, including the time a person spends checking the output.

What an employee costs per hour in your EU country

The salary is not the cost. Your cost also includes employer social contributions, paid leave, sick pay and training. Eurostat publishes this every year as the hourly labour cost: everything an employer pays, divided by the hours people actually work. It covers companies with 10 or more employees, across the whole economy except agriculture and public administration.

Because it is per hour worked, paid leave is already inside the figure. EU law guarantees every worker at least four weeks of paid annual leave (Working Time Directive 2003/88/EC, Article 7), and many countries grant more. You pay for those weeks, but nobody enters orders during them.

Source: Eurostat, labour cost levels (lc_lci_lev), 2025, data updated 23 April 2026. The last column is the hourly cost × 15 hours × 52 weeks.

Across the EU, 24.8% of labour cost in 2025 was not wages, mostly employer social contributions. That share differs a lot between countries, because countries split social contributions differently. The total hourly cost is the number to compare.

The figure has two limits. First, it is an average across all jobs, so an admin or order-entry role may cost less than the average and a specialist more. If you already employ someone in a similar role, use their real cost instead: annual gross pay plus employer contributions, divided by the hours they actually work.

Second, it leaves out management time: supervision, cover, and explaining the same exception twice. Add that as its own line if it is real for you.

For a floor, look at the statutory minimum wage. On 1 January 2026, 22 of the 27 EU countries had one, from €620 a month in Bulgaria to €2,704 in Luxembourg (Eurostat, gross, before employee deductions). Germany's was €2,343, Slovenia's €1,278 and Croatia's €1,050. Denmark, Italy, Austria, Finland and Sweden have no national minimum wage.

Employer contributions come on top of the minimum wage, so the minimum wage is never your full cost.

What your weekly process hours are worth

Multiply the loaded hourly rate by the weekly hours, then by 52. We use the EU average of €34.90 for the worked example. Replace it with your country's figure or your own payroll figure.

• 15 hours × €34.90 = €523.50 a week.
• €523.50 × 52 = €27,222 a year, or €2,268.50 a month.

That is what the process costs you today, and roughly what you would pay a hire to absorb those hours. If salaried staff already do the work, your extra cash cost may be lower. Keep the full rate anyway: those hours could be spent on other work.

What an automation pilot costs after launch

An automation pilot has three cost lines:

• Setup: the one-off work of mapping the workflow, connecting your sales platform and accounting software, and testing on real orders.
• Running: hosting, tool and model fees, maintenance, and handling failures when a software vendor changes a field or an integration.
• Checking: the time a person spends reviewing output before it counts as done.

Check that a quote accounts for the third line. Automation with approval gates does not remove the person. It moves the person from doing the task to checking it.

A verification gate is a fixed rule, not a model's judgement, that runs before an output can go further. In order entry, a person still approves sending an invoice or posting a payment.

Price the checking like any other hours. Two hours a week at €34.90 is €3,629.60 a year, or €302.47 a month.

We don't publish price ranges. Our AI automation is priced per workflow after an audit: a pilot first, then rollout, priced by complexity and hours removed (see the FAQ). So instead of an example quote, the worksheet below tells you the most a quote can cost before it stops paying back.

When you ask any provider for a quote, ask for these lines separately:

• Setup, one-off.
• Hosting and tool fees, monthly.
• Maintenance and change requests: included hours, response time and the price per change.
• Exception handling: who handles an order the automation cannot process, and at what cost.
• Monitoring: whether you get access to the error logs.
• Checking: the hours your team will spend approving output.

The break-even worksheet: find your payback ceiling

Work through five steps.

1. Take your monthly process cost from the section above. EU example: €2,268.50.
2. Subtract the monthly checking cost. EU example: €2,268.50 − €302.47 = €1,966.03. This is what the automation saves you each month before its own running costs.
3. Choose the payback period you would accept, for example 12 months.
4. Multiply: €1,966.03 × 12 ≈ €23,592.40. This is your payback ceiling. The setup cost plus 12 months of running costs must stay below it.
5. Compare each quote with the ceiling. Setup plus running costs below it: the automation pays back within your period. Above it: hiring, or a smaller first step, wins on cost.

All three columns use 15 process hours and 2 checking hours a week.

The ceiling moves with your inputs. More weekly hours or a higher hourly cost raise it. More checking time lowers it. If checking takes as long as doing the work, there is nothing to save.

The worksheet gives you an estimate, not a guarantee. It tells you when the arithmetic favours automation if the process is ready. It cannot tell you whether it is ready. The next section does.

Before you automate: four preconditions

Four things need to be true before the break-even month means anything. Our post on when not to automate a business process turns them into a one-day test.

Stable rules. Most orders follow the same entry and invoicing rules. If every order needs a judgement call, you would be automating judgement, not a process.

Clean data. The fields coming from your sales platform and spreadsheets are consistent and complete. Duplicates and missing fields turn into expensive exceptions later.

A known exception rate. You know roughly how many orders need a person, what triggers them, and who handles them. An unknown exception rate is a reason to measure first, not to build.

A team that accepts approval gates. People review output before anything irreversible happens, such as an invoice going to a customer.

Even when all four hold, the person does not disappear. In every build we deliver, a person on your team approves anything that leaves the building. We add verification checks where mistakes are expensive, and budget ceilings to multi-step AI pipelines. The software flags the exceptions, and a person handles the difficult customers.

After launch, someone has to watch it: our runbook for monitoring and maintaining automations lists what to check weekly.

Time to value: recruiting versus automating

A hire has a clock that starts before the first payslip: the job ad, screening, interviews, the notice period at the candidate's current employer, and weeks of onboarding before the person works at full speed. How long this takes depends on the role and your labour market, so treat it as your own planning assumption.

An automation pilot has a different clock. Ours starts with an audit of how the work moves through your company. Then we pick the steps AI can do reliably and the steps that stay with a person, and build a pilot on real work before any rollout.

Our usual sequence is one workflow mapped in week 1 and a pilot running on real work in weeks 2–4. That is a sequence, not a promise for order entry: we have no order-entry case study, so we don't quote a delivery time for one.

Questions to ask any automation provider

Get written answers before you sign. They work for any provider, including us.

• What is the contract length, and how do we exit if the pilot fails?
• Who owns the code and the data at the end?
• What happens when our accounting software, sales platform or ERP changes an integration?
• Which actions still need human approval, and how is the approval step built?
• Can we start with one process and switch it off without a long commitment?
• How are errors logged, corrected and kept from happening again?

The FAQ answers how each service is priced and whether you need a long-term contract.

What to do next

If you now have a monthly process cost, a payback ceiling and an honest answer on the four preconditions, test them on one real process. Our AI automation service starts with exactly that audit. Book a business-process review and we'll map the process with you, including telling you plainly if it is not a good candidate for automation.