How do you work out a person's hourly cost?
A reasonable approximation is to take the employee's total annual cost — salary plus benefits and payroll burden — and divide it by about 1,900 working hours a year. That number usually lands between 40% and 55% above the nominal hourly salary.
If you want a quick estimate, take the monthly salary divided by 160 and add 45%. The point is not decimal accuracy: it is not using take-home pay, which systematically understates the real cost.
Why does the result usually look exaggerated?
Because repetitive work is felt in fragments and paid for in aggregate. Nobody feels that copying data between two systems is a problem when it takes fifteen minutes; across a year, those fifteen minutes a day for three people are more than 180 hours.
That is exactly why the most profitable automations are not the spectacular ones but the boring ones: they go after tasks nobody defends because nobody had measured them.
What share of a task can really be automated?
It depends on the kind of work. Tasks that shuffle data between systems automate almost completely. Tasks that need judgment in ambiguous cases automate partially: the system handles most of it and sends the rest to a person.
As a conservative reference for a first estimate, 60% is a prudent assumption on data tasks and 30% on tasks requiring judgment. Better to come in low: an automation that pays off more than expected does not produce awkward conversations.
And if the saving does not turn into a smaller payroll?
It usually does not, and that is worth saying plainly. The return on automating is rarely laying people off: it is the same people spending those hours on work that does need judgment.
Which is why the figure this calculator produces reads better as capacity freed than as an accounting saving. At Articod we use it to prioritize what to automate first, not to justify cuts.