RPA and Process Mining

Software AG partners up with Automation Anywhere to provide Process Mining to AA's customers — continuing a recent trend.

February 14, 2021
893 words (5 min read)
Automation Anywhere
Process Mining

Software AG + Automation Anywhere

Software AG announced this week that they are partnering up with Automation Anywhere to provide process mining (and monitoring) capabilities to the latter’s customers and RPA to the former’s.

In itself, this sort of partnership means very little in my experience. Partnerships are not created by signing agreements but by employees from the two companies going into joint customer opportunities, formally or informally. Supposedly all RPA vendors partner with all Process Mining vendors and vice-versa, but in reality next to no collaboration seems to be going on.

The main utility, unfortunately, is to provide a way for vendors to defend against accusations by customers that “they do not have capability X” during the procurement process so the customer can tick their little boxes on the RFI evaluation form.You can probably infer my opinion about the practice and Requests for Information in general

Process Mining + RPA

So why do I care about this one, then? I don’t, really, except that it highlights an interesting trend: UiPath acquired ProcessGold in 2019 (it’s now UiPath Process Mining), Microsoft partnered with PAF to enhance PowerAutomate with PM, SAP used to partner with Celonis and now acquired Signavio, so it’s only natural that AA wants to avoid being left out in the cold.Why are all these process mining vendors German, by the way? Celonis is from Munich, Software AG and PAF from Darmstadt.

It seems, therefore, that all major automation vendors consider process mining to be an integral part of automation’s value proposition. Why would that be so and what does Process Mining actually bring to the table?

What is Process Mining?

While a more in-depth treatment of this topic has to wait for a longer post, very briefly, process mining reads timestamp data from backend systems and logs (e.g. SAP) in order to construct a so-called event log. These timestamps have to be flagged with a case id, e.g. an invoice number, to be useful.

The resulting process trace is then visualized and enriched with additional metrics to provide insights into the actual process execution and — crucially — the duration of the different steps. This provides an empirical view on how the business process is actually done in practice, rather than an inevitably idealized notion of it.

Show Me the Money

Customers use process mining to provide value in several ways:

  • Knowledge is power: knowing how your process actually works gives you a basis to improve it. PM helps identify bottlenecks, detect policy violations, find waste such as rework, etc.
  • Problem zones: when analyzing things like invoices, you will often find that a small number of suppliers or geographies cause a large percentage of issues. PM provides evidence that allows you to address these shortcomings.
  • Automation potential: most relevantly for us, most business processes are chain-like, i.e. their throughput is limited by their slowest step. Thus, knowing the bottlenecks allows you to use automation much more strategically.
  • Monitoring: more recently, there seem to be some efforts to make process mining more real-time focused than backward-looking, by providing abilities to trigger robots or send some information to a person if something goes wrong. PM can be a useful tool for this because of its case-based perspectiveFor example, you can monitor payment terms and proactively emit a reminder if an invoice is likely to miss the payment target


So is process mining a panacea? Well, no, obviously not. The main problem it has is that tasks which are not logged in your backend system are invisible for PM. This is particularly the case for manual steps, but you might be surprised by how rarely systems store a useful change history, too.

This can lead to missing the most important or time-consuming parts of the process. Some vendors (at least UiPath and Kryon) attempt to address this by adding task mining to the mix (i.e. monitoring what users actually do on their computer with RPA technology) to close the gap. Personally, I am skeptical of this approach as the data is so under-specified that this seems highly unlikely to deliver on its promises without significant manual input by a process expert

A second shortcoming of process mining technology is cost, especially the high per-process cost of implementation. While vendors are attempting to address this by adding standard processes to the mix, given the real-world variation of processes between different companies, you should still expect a pretty significant effort for every single process you investigate.

What’s worse, you need a pretty specialized kind of expert to do the implementation: people who are technically savvy but also have a good understanding of the business process or can acquire it quickly. Whenever you move into the intersection part of a Venn diagram, this cuts down the number of available resources quickly. Most consultants are, unfortunately, not good enough on either the technical or on the process side of things to really deliver the goods — and those who are, are in short supply.

Wrap up

That’s it from me for this week. See you next time!

© 2021, Stefan Reutter