Zbigniew Misiak runs a poll every year in his blog BPM Tips about the BPM skills that are required. This year was a challenge for me, as I see a big change happening as we move from bespoke BPM application to transition ultimately to learning based applications. Find my response below.
BPM Skills in 2020
This has been particularly difficult for me to put together this year, because really many things have changed outside of the BPM field, that actually have profound, but as yet unrealized effects on the field.
When we look at BPM from 2003 – 2018 we see a focus on process modeling, and other kinds of modeling. Decision modeling is the latest in this trend. What is a process model? It is after all a depiction of what the people in the organization think the preferred way of handling a business problem is. This has always been negatively affected by the fact that much of this knowledge is tacit, and not consciously known by the workers. So interviews work to a limited extent, but the real process is more difficult to suss out. The next technique was one of Agile trial and continuous improvement: approximate the process as best you can, measure performance, and incrementally improve the process. But all of this is really traditional development: a programmer take the logic and codes it in a process model, in web services, or in glue code.
Today we are in the age of Alpha Go and learning software that is trained, and not programmed. These game-playing computers did well in extremely complex games like Go where the number of rules is small, but the possible combinations very high. The technique is likely to be even more useful when the exact rules are not known, like the rules of a business process. AI should watch the behaviors of the individuals, and be able to draw up rules of engagement of the business processes. Those rules may be completely opaque to the workers, managers, and programmers. That does not matter. What matters is that the rules work. Learning systems are likely to be able to route work to people far more efficiently than any externally applied process logic.
What this really means is that the traditional skill of modeling a process, and programming applications, is no longer relevant. Let me temper that: these will still be used to create BPM applications for another 10 years or so, but the demand for this will drop quickly. Instead, we will find that forward thinking organizations will deploy learning systems to watch the organization at work, to automatically identify processes that are stuck, and to proactively route them forward for completion. The process might be difference every single time, but that does not matter: the goal is efficiency of the organization, not regularity of the process.
I have attempted to explain all this in my book “Beyond the Business Process Model” where I outline not only the trend toward learning systems replacing bespoke BPM applications, but also outline the parameters that would be needed to make it all work. I must humbly warn you: this idea has not been very well received in the industry, and there has really not been much uptake in readership of the book.
I think that process modeling and decision modeling are currently at their maturity. It will take a decade to swap out the current bespoke application approach, to a learning approach. The skills that you will need for that newer technology will be data science and deep learning. We are seeing this same trend in many areas that were traditionally implemented using standard programming techniques. We all need to learn new skills, and I can promise, it will be a fascinating ride.