To RPA or RDA - that is not the question! Rather: How do you find a valid RPA or RDA case 🤖


It is the set of the sails, not the direction of the wind that determines which way we will go.

LinkedIn version here.

We all love robots.

What is just annoying though: the time spent on finding valid RPA/RDA use cases.

So a valid robot use case should usually answer ALL of the following questions somewhat positive:

Some real examples could be:

  1. I have seen RPA use cases to automate functions in SAP, like “create billing invoices”. But hey, isn’t it a standard function? Yes, but it only works, if all data (especially master data) are right. Fully automation by using standard functionality is always better than automating via a robot.
  2. This is an obvious point that needs no further commenting.
  3. Doing thousands of postings every month - possible RPA use case. Doing thousands of postings at year end… maybe not. What would the robot do for the other 11 months? How can it be, that we have a known problem every year at the end of the year and never been able to fix it, but now that we have robots, we suddenly think we should/could/must? It starts with a mindset.
  4. Our robots can handle IF and ELSE - even multiple IFs and ELSEs. But not sometimes IF and possible ELSE, it depends… Any human can, though.
  5. Sorry, e-mail is not really structured. And information from a phone call? Or a handwritten note? Good luck.
  6. Sorting fruit is easy. Apples, oranges, bananas. Even Asian pears that sometimes look like an apple. Throw in a cucumber and see what happens. And trust me, as much effort and care we do take for point 5, it still happens. So maybe we should just make sure that the process does not require so many exceptions! Yes, we definitely should. But that means we are indeed focusing on the process improvement - and not necessarily on the RPA aspect as a mean to an end!

But hey… too tedious work!
I agree - but it is not a matter of skipping tedious, it is a matter of making tedious easy and quick.

And this is what Process Mining can and has done for us.

We created a very simply analysis:

a) we show the current robot activities and also where they are used for best practice sharing or just to give a heads up where you could start:

b) On a simple X/Y Axis we should the % and the absolute numbers --> the assumption is that high manual rate and high number of absolute occurrences is a high potential:

c) as you make more meaningful selections (aka adjusting the filter criteria like specific activity, country, BU, etc.) the Potential is moving around like here:

–> so in the above example Change Price would be a good start to start the discussion. In many cases you probably want to avoid rather to fix, though (where possible).

Using this approach is also a good way to quickly go from gut feeling to facts and then validate these facts into the right context --> by customer, by time period (does it happen only at quarter end?).

As I mentioned many times - possibilities are limitless!