The question nearly always comes up the same way. The work has grown, the office is behind, and the first idea is to add a person. Before you do, look at what that person would really spend the day on, because a share of those hours asks for no skill at all, only patience.
First of all, they aren't the same thing
You need a person when a decision has to be made, when the case is odd, when somebody has to talk to a customer or take responsibility. Automation covers the opposite end, the steps that repeat identically and follow rules you could already write on a sheet of paper: take the number from the invoice, look it up in the business software, if it doesn't match, stop.
Look at the day, not the job title
The job ad is nearly always generous, the real day is another matter. Take the role you have in mind and break it into the actual tasks of one week, then note next to each one how long it takes and whether the way you do it changes from case to case or not.
The tasks that come out identical every time are the ones to put on the table. Copying data from an Excel sheet into the business software, downloading receipts from a portal, reading PDFs that are all alike and carrying over four fields, checking at the end of the day that everything got entered.
The sum on the hours, done properly
The basic calculation is simple. Multiply the people who do that work by the hours a day each, then by the days per month, then by the hourly cost to the company. The hourly cost is the full one, not take-home pay divided by hours, contributions and everything else go in there.
Here's an example made up just to show the method. Say two people, two hours a day each, twenty days a month. That's eighty hours a month, close to a thousand hours a year, and from there you work out the cost with your own figure. The number can be a surprise, because those hours never arrive as an invoice, they're scattered through the days.
Then add the items almost nobody counts. The time spent checking, the errors to fix, the interruptions to hunt down a figure that doesn't add up, training whoever joins, and the backlog that piles up when somebody is on holiday or off sick.
How much of it can really be automated
This calls for honesty, it's almost never all of it. A typical process has a large part that always follows the same steps and a tail of odd cases somebody has to look at. Be cautious with the estimate, if you think eighty percent of the time can go, work on sixty and see whether the sum still holds.
A useful sign. If the person who does the job can explain the procedure to you in five minutes without ever saying it depends, that part can be automated. If instead she says it depends every other sentence, there's human judgment in there and it belongs to a person.
What differs between the two roads
An extra person gives you flexibility. She understands the new cases, talks to customers, covers things you hadn't foreseen when you hired her, and she costs you every month she stays. An automated system understands nothing you haven't explained to it, but it takes the peaks without asking for overtime, works at night if that helps, and when volumes double it doesn't ask for twice the time.
There's a difference in timing too. You hire a person and then she has to be trained, an automation has to be built and tested, and either way the gain doesn't arrive the next day.
When hiring is still the right call
If the work changes shape every week, if there's a lot of contact with customers, if the decisions carry weight and need a name attached to them, then you hire the person and that's that, there's no calculation to do. Same if the volumes are small, ten cases a month don't justify building a system around them, you do them by hand and you sleep fine.
The interesting case is a different one. The volume is high, the work is always the same, and you're about to hire somebody to do it all day. There, do the sum first.
The third road, the one in the middle
The best choice is rarely all on one side. Take the repetitive hours off the workload, look at what's left, and then decide whether you still need an extra person or whether the ones you have cover the work at a calmer pace. It also happens that you need the person anyway, but for something that brings value instead of re-entering data.
You aren't pushing anybody out. You're only stopping paying a skilled person to copy numbers from one screen to another, which is a waste for you and a bore for her.
Where to go from here
If the sum comes out small, you've settled the question in half an hour and you hire with a clear head. If it comes out large, before you post the ad look at how much of those hours a system could take, and compare that saving with what it costs to build it and keep it running.
If you want a second opinion on your case, we need to see the program you use, the steps you take every time and how many operations pass through your hands in a month, because those are the things that say whether a process is a fit or not.