Key takeaways
- Automation only pays when the time it removes is worth more than it costs to run. This calculator finds the monthly cost at which that stops being true.
- The number that decides the outcome is usually residual time — the minutes a person still spends reviewing or approving each automated run. Set it honestly and many marginal cases stop being marginal.
- Coverage matters more than speed. A process automated for 70% of runs saves roughly 70% of the labour, no matter how fast the automated path is.
- Treat the result as a threshold, not a forecast. It tells you what has to be true for the automation to be worth it, which is a far more useful thing to argue about than a projected saving.
Estimate one process
The defaults are editable assumptions, not benchmarks. Replace every one of them with your own figures — the result is only as good as what you put in.
- Hours saved per month
- 30.3
- Gross labour value
- €1,365
- Net benefit per month
- €875
- Return on investment
- 178.6%
- Payback
- 1.0 months
- Break-even monthly cost
- €1,365
At these inputs the automation pays for itself as long as it costs less than €1,365 per month. You are budgeting €490, leaving €875 of value a month, and the 20 hours of setup pay back in about 1.0 months.
How this is calculated
The whole calculation is seven lines of arithmetic, and hiding them would make this page a lead magnet rather than a tool, so here it is in full.
minutes saved per automated run = max(0, manual minutes − residual minutes)automated runs per month = runs per month × coverage ÷ 100hours saved per month = automated runs per month × minutes saved per automated run ÷ 60gross labour value per month = hours saved per month × hourly costnet benefit per month = gross labour value per month − monthly automation costROI = net benefit per month ÷ monthly automation cost × 100payback = (setup hours × hourly cost) ÷ net benefit per monthbreak-even monthly cost = gross labour value per month
Two results deserve more attention than the headline ROI figure. Break-even monthly cost is the gross value of the time saved: spend more than that and you are losing money regardless of how good the automation is. Payback charges the one-off setup effort at the same hourly rate as the work, and tells you how many months of net benefit it takes to recover it. When the automation does not pay for itself, payback is reported as no payback rather than as a large number.
Where the defaults come from
| Input | Default | Why this starting point |
|---|---|---|
| Runs per month | 200 | Roughly ten a working day: frequent enough that the process is worth pinning down, which is the case where this question comes up. |
| Manual minutes per run | 15 | A short, boring, multi-system task. Long enough to hurt in aggregate, short enough that nobody has prioritised fixing it. |
| Fully loaded hourly cost | €45 | Salary plus employer costs divided by working hours. Use your own figure; this one is deliberately mid-range and will be wrong for you. |
| Automation coverage | 70% | Assumes a meaningful minority of runs still need a person. Assuming 100% is the most common way these estimates go wrong. |
| Residual minutes per automated run | 2 | Reviewing and approving is not free. This is the input that most often turns an attractive estimate into a marginal one. |
| Monthly automation cost | €490 | A placeholder in the region of a platform subscription. See pricing for what actually drives this. |
| One-time setup effort | 20h | The time your team spends teaching and checking the process before it runs on its own. |
A worked example
A finance assistant codes 200 supplier invoices a month. Each one takes about 15 minutes of reading, matching and keying at a fully loaded cost of €45 an hour. The team expects 70% to go through without intervention, with two minutes of review on each of those.
That saves 13 minutes on each of 140 automated runs — about 30.3 hours a month, worth €1,365. Against €490 of monthly automation cost, the net benefit is €875 a month, an ROI of roughly 179%, and the 20 hours of setup pay back in about a month.
The instructive part is the threshold: this process stops being worth automating above €1,365 a month. And if coverage turns out to be 40% rather than 70%, the value of the time saved falls to €780 — still positive, but a much less comfortable case. Those figures are illustrative, not measured results.
How to read the result
Treat the output as a threshold rather than a forecast. It is not a claim about what you will save; it is a statement of what would have to be true for the automation to be worth it. That is the more useful thing to argue about in a room, because the argument moves to the coverage assumption, which is where it belongs.
It is also worth remembering what this calculation does not reward. Google's site reliability engineers describe toil as work that is manual, repetitive, automatable, produces no enduring value, and grows in direct proportion to the business. The last property is the one that does not show up in a monthly figure: the labour cost of an unautomated process rises as you grow, while the automation cost largely does not.
Limitations and when not to use this
- This counts labour time only. It ignores the value of fewer errors, faster turnaround, work that stops being deferred, and the cost of the mistakes an automation can make. Those are real, and none of them are in this arithmetic.
- It assumes the process is stable enough to automate. A process that changes shape every month will spend your setup hours repeatedly, and this calculation charges them once.
- Coverage is an estimate you are making about work you have not automated yet, and it is the input people are most optimistic about. If the result is only attractive above 80% coverage, treat that as a warning.
- This is an educational estimate, not financial advice, and it is not a quote. It makes no claim about what any particular product will achieve for your process.
Sources
- Eliminating Toil — Google, The Site Reliability Workbook Accessed 24 July 2026
- Embracing Risk — Google, Site Reliability Engineering Accessed 24 July 2026
Talk through your numbers
Bring one process and the figures you put in here. Thirty minutes is usually enough to work out whether the coverage assumption holds.
Talk through your numbersUli Prantz
Builds and operates all-agents
Uli Prantz builds all-agents, the process-automation platform this site documents. He writes about the operational side of automating recurring business work: where deterministic code beats model judgment, where it does not, and where a human still has to approve.