A most interesting talk at BPMN2012 today was about case management, presented by Volker Gruhn on “Managing & Tracing the Traversals of Process Clounds with Templates, Agendas and Artifacts” from a short paper he did with his collegues from Essen: Marian Benner, Matthias Book, Tobias Brückmann, Thomas Richter and Sema Seyhan.  They confirm all the design principles of ACM.

The concept “Open Process Clouds” has nothing to do with cloud computing, but rather parts of a process that can not be presented by a traditional process. They use a little cloud symbol in the process definition to denote the ambiguous, unpredictable behavior of what people do there. This is precisely what we have been talking about in ACM as an unpredictable process, but his terminology is a little different.

“many critical business processes lack a predefined process model, but depend upon the experience and expertise of case managers who work with large amounts of unstructured data to make decisions.”

He gave two examples of these:

• reinsurance underwriting – an insurance company will transfer parts of their risks to another party, and this behavior looks like bargaining. There is no strict sequence of activities, and this is not supported well in workflow manager or BPM.
• health insurance rehabilitation – this is a long term control process with many people involved, and it is difficult to say ahead of time exactly what is needed, and depends upon so many details that are not clear from the beginning.

The important thing to notice is that these are not trivial activities, but instead important, critical business operations. Even though these are so valuable, for these business activities it is impractical or impossible to specify fixed interaction patterns. There are a number of different reasons for this:

• Either the pattern is too complex or too rarely executed to fix the exact pattern. These business activities happen from time to time, but not on a daily basis. For example in injury rehabilitation something may be needed one day, and then not again for several months.
• The business activity may be too interactive, with many people being involved, and we can’t predict the exact timing of when people are available.
• The understanding by the expert may be fuzzy and non-algorithmic. The experts know what to do but can not explain it.

In all of these cases, the expert competence and knowledge remain essential for such tasks!

The example in the paper is an injury rehabiitation control process. The first step “generate healing plan” was not really defiend how to do that and so is represented as a little cloud. However, this step is required to generate a “Healing plan”. The second part of the process is to monitor the plan, and again it is impossible to pre-define exactly what has to happen there. After that is completed, an evaluation report is written.

An interesting thing to note is that while the experts have an idea about the artifacts they should produce, and possible the databases they should access, they often don’t know the details about the sequences of action to achieve that. Others interacting with the case don’t care at all about the activities or the order they are done, but do care about the artifacts and structure. This is consistent with the ACM idea that we should be focussing on data and artifacts more than on the process.

Agenda  – They propose the term “agenda” for the collection of tasks that need to be done. It is a simple list of goals to be accomplished, and there is nesting to allow sub-goals within goals. Also, there is a required artificat that must be input to this process cloud, as well as the artifacts that will be produced.

“Process clouds do not have a fixed sequence of steps, but usually has some guidance on what should be accomplished.”

Templates – To enable learning and evolution they formed “templates” with proposed artifacts and agenda. It is necessary to ba able to modify the agenda on an instance by instance basis, possibly adding tasks or removing tasks to fit the specific case. They also have the opportunity for a case manager to start with a “clean slate” which is simply an empty list from not using a template. Also, they have a way for a case manager to “propose” a template for next time they do this again, that is, correcting what they felt was a mistake so that next time they can do it better.

One of the biggest advantages they see is the accumulation of traces from previous cases. These can be analyzed with process mining techniques, and will give you information to improve the handling of the cases in the future.

Summary: I like this paper because it puts very good words around the issues that the ACM people have been talking about: important business activities can not be modeled and supported with a BPM approach. Instead of a process, they use a very simple concept of an “agenda” which can be modified by the case manager at any time. The focus needs to be more on artifacts, and less on task/process. This is clearly the right idea for knowledge worker cases.