IT tools allow us to automate all kind of tasks with the aim of reducing time and errors. But not everything should be automated. An inefficient and complex process can be sped up using some kind of automation, but it will still be inefficient, and because of its complexity, probably the cost (in time, in validations, and in money) of automating the process will be greater than expected.

It is much better to improve your business processes and think about automation once they are good and efficient.

You should also think about the importance of the process and about how often that process is done. A process that is done dozens of times per day is probably worth automating, while a process that is done once a year may not be the best candidate for automation.

The process in our example has to be clearly improved before any kind of automation is applied to it. We cannot just take it as it is and automate only some of the activities.

The activity that probably takes the longest is going through all the purchase invoices and determining whether they are attributable to a specific service. The determining part probably cannot be automated. We need a human to take this decision. What about going through all the invoices? Yes, maybe we could get a robot to go through our physical archives, get all the invoices of a provided list of vendors, present them to us one by one, and then archive the paper again.

This sounds cool, but it makes no sense at all. We definitely have to change and improve this process before we can automate anything.