Calculation method

This page explains how the simulator gets its results. The calculation is simple: it relies on a few assumptions described below.

What the simulator sets out to find

A repetitive task costs people time. Automating that task with AI requires an investment, in the form of development days. The simulator answers a single question: once the automation is live, how many months does it take for the gains to cover the investment? It does so twice, once in time (the Time view) and once in money (the Cost view).

Time saved

We start from the task's duration, entered in minutes or hours (minutes are converted to hours by dividing by 60), and its frequency: a number of times per day, per week or per month. The number of working days per year is entered in the Time view. It is 228 by default, an average for an employee in France: the 261 weekdays of a year, minus about 8 public holidays and 5 weeks of leave. A daily task counts as many occurrences per year as there are working days. A weekly task counts as that number divided by 5 (45,6 weeks by default), and a monthly task as 12.

We multiply the task's duration by the number of occurrences per year to get hours per person per year. We then divide those hours by the length of the working day (7 h by default, adjustable) to convert them into days. Finally, we multiply by the number of people involved to get the days saved per year for the whole group, and divide by 12 to get the days saved per month. The “time saved per month once the task is automated” shown in the Time view is this time, expressed in days and hours, for all people combined, over a month or a year depending on the Month / Year toggle.

A key assumption: AI replaces 100% of the task. If the task takes 30 minutes today, the simulator assumes it takes zero minutes tomorrow.

Time invested and payback in time

The time invested is the number of development days multiplied by the number of developers. Developers are assumed to work with an AI coding assistant: the number of days entered should therefore be that of a developer equipped with this tool. The simulator does not adjust it.

Payback in time is the number of days invested divided by the number of days saved per month. The result is a duration counted from the moment the solution goes live. If a go-live date is entered, the simulator derives an approximate profitability date from it: it adds the duration to that date, in calendar months, and moves a date that falls on a weekend to the following Monday.

The calculation in money

Each person involved has a gross annual salary. A day is worth this salary divided by the number of working days per year. The annual gain is, for each person, the value of a day multiplied by their days saved per year, all added together. From 10 people onwards, only an average salary is requested, applied to all the days saved. The monthly gain is the annual gain divided by 12.

Recurring costs (subscription, hosting, maintenance) are optional and zero by default. They only count in the Cost view: the Time view ignores them. If entered, they are subtracted from the monthly gain to give the net gain per month. If this net gain is zero or negative, the simulator shows “Never profitable” rather than an absurd duration.

The development cost is the number of days invested multiplied by the value of a developer's day, that is, their gross annual salary divided by the number of working days per year. Payback in cost is the development cost divided by the net gain per month.

What the calculation does not take into account

The simulator is optimistic on some points and cautious on others.

It ignores human review time: an answer produced by AI often needs checking, and this time is not counted, since the task is assumed to be 100% replaced. It uses gross salary, without employer contributions, which underestimates the real cost of a working day, and therefore both the gain and the development cost. It does not model any learning curve: users are assumed to be efficient from day one. Nor does it model the adoption rate: everyone is assumed to use the solution, all the time. Finally, recurring costs are optional and zero until they are entered, which flatters the payback shown.

The Month / Year toggle only changes the display: it changes the unit of the payback, the net gain and the time saved, as well as the chart's scale, never the calculated value. In the Year view, payback is expressed in years, with the net gain per year and the time saved per year; the development cost does not change, since it does not depend on any period.

A step-by-step example

Take 5 people who each earn 50 000 € per year, and who spend 30 minutes, 1 time per day, on a task. The working day is 7 h and the year has 228 working days. 2 developers earning 60 000 € per year spend 10 days each on it with an AI coding assistant, and there are 200 € of recurring costs per month.

Each person saves 16,29 days per year. For the 5 people, that makes 81,43 days per year, or 6,79 days per month. The investment is 20 development days (2 developers for 10 days each). Payback in time is therefore 20 ÷ 6,79, or 2,9 months.

In money, the gross gain is 17 857 € per year, or 1 488 € per month. Once the 200 € of recurring costs are deducted, the net gain is 1 288 € per month. Development costs 5 263 €. Payback in cost is therefore 5 263 € ÷ 1 288 €, or 4,1 months.