How to calculate the ROI of an automation

Automation for automation's sake is pointless. An automation is an investment, and like any investment, it should be calculated before you commit. The good news: the maths is simple, as long as you are honest with the numbers. Here is the formula, a complete worked example, the costs people forget, and the cases where the right decision is not to automate.

The basic formula

Three elements are enough:

Monthly gain: hours saved per month x the fully loaded hourly cost of the person concerned.
Project cost: designing and deploying the automation, usually a fixed price.
Running costs: AI API calls, tool subscriptions, maintenance, every month.

The profitability calculation follows:

Break-even point (in months) = project cost divided by (monthly gain minus monthly running costs).

Under 12 months, the project is generally solid. Under 6 months, it is an excellent candidate.

Estimating time saved without fooling yourself

This is the most important number, and the most commonly overestimated. Three rules:

Measure, do not guess. One week of actually tracking the task (how often it is done, how long it takes each time) beats any hallway estimate.
Apply a discount. An automation never captures 100 percent of the theoretical time: edge cases, supervision, exceptions. Keep 70 to 80 percent of the measured time.
Use the fully loaded hourly cost, not the net salary: gross salary plus employer contributions. For an administrative or sales role in a small business, expect roughly 25 to 40 € per hour.

A complete worked example

A services company automates its shared inbox: message classification, drafted replies, CRM updates.

The inputs:

time measured on the task: 6 hours per week;
after a prudence discount, time retained: 5 hours per week;
fully loaded hourly cost: 30 €;
project cost: 3,500 € fixed price;
running costs: about 30 € per month in API calls and 120 € per month for maintenance and supervision, so 150 € per month.

The calculation:

monthly gain: 5 h x 4.33 weeks x 30 € = roughly 650 € per month;
net monthly gain: 650 - 150 = 500 € per month;
break-even point: 3,500 / 500 = 7 months;
year one: around 2,500 € of net gain once the project has paid for itself;
year two: around 6,000 € of net gain.

Want to stress-test the case? Run the same calculation with the low estimate: 4 hours per week instead of 5. The monthly gain drops to around 520 €, the net gain to 370 €, and the break-even point moves to 9 or 10 months. Still acceptable. If a project only works with the optimistic numbers, it is not a solid project.

An honest ROI on this kind of project is measured in months, not days. Be wary of promises of instant profitability: they usually ignore running costs or overestimate the time saved. Our case studies show the orders of magnitude we see on real projects.

The hidden costs to include

The project fee is not the only cost to plan for:

Your internal time: scoping, providing access to tools, testing on real cases. Count a few hours on your side, especially early on.
Training and adoption: an automation the team works around returns nothing.
Evolving maintenance: when a tool changes its API or your process evolves, the workflow has to follow.
Supervision: someone has to watch the alerts and handle exceptions. Not much time, but not zero.
Variable API costs: if the processed volume doubles, the AI API bill follows. Worth monitoring, even though it usually stays modest, a few dozen euros per month for most SMB use cases.

Indirect gains: real, but keep them out of the calculation

Fewer data entry errors, no forgotten follow-ups, faster replies to customers, teams refocused on higher-value work: these benefits are real and often weigh more than the time saved. But they are hard to quantify honestly.

Our recommendation: base the decision on the time x hourly cost calculation alone. If the project holds up on that criterion, the indirect gains are a bonus, not a justification.

When NOT to automate

The calculation is also there to say no. Some clear signals:

Volume too low. A task that takes 2 hours per month will almost never justify a custom project: the break-even point stretches into years.
Unstable process. If the way of working changes every month, the automation will spend its life in maintenance. Stabilise first, automate second.
Undefined process. If nobody can describe the steps precisely, there is nothing to automate. Automating vagueness just produces vagueness, faster.
Closed tools. Without an API or an export, integration costs more than the gain. Check this before going any further.

In those cases, it is often better to start with a simpler automation on another process with more volume.

The takeaway

An automation ROI calculation fits in three lines: time saved x hourly cost, minus running costs, compared to the project cost. Be conservative on the time saved, exhaustive on the costs, and accept that a good project pays for itself in a few months rather than a few days.

That is exactly the calculation we run with you during the free 30-minute assessment: your processes, your volumes, your costs, and a numbers-backed estimate before any commitment. See how our method works.