I have a curiosity / thought sparked from this great article that is only adjacent to the topic at hand, it has nothing to do with Graph engineering and AI. I wonder why BPMN had some moderate amount of success in the 2010s, while other PML attempts failed. I was running the engineering team of a company called Appway, which built a BPMN tool for modeling and executing banking processes. We had a decent amount of success which lead to a great exit in 2021, and the competitive landscape was fierce. Companies like Appian, Pegasystems and the OS Flowable are still amongst the largest players in the space, alongside Mendix, Outsystems, Camunda and many more smaller companies.
The issues highlighted in the "what went wrong" chapter of this post are similar to what we faced when modeling banking processes with BPMN: the promise of the language was to lower the incidental complexity of coding them in C++/Java/C# and allow businesses to focus on the inherent complexity of the process at hand. The reality was that the formalism required to turn the BPMN model into an executable, required going down to the details and model each and every variable passing, logical condition and decision element. These complex models required the skills of SW Engineers to build and maintain, that at that point failed the promise of "getting the business closer to the application". However some of the commercial products I mentioned above are still quite successful in the enterprise landscape; I thought AI would deliver the final blow, but it seems like it didn't from the company performance of the aforementioned players (which aren't thriving, but they aren't failing either). What made the difference in BPMN / Workflow automation vs PML at large?
BPMN is simpler and more focused. Therefore, more effective. UML has some process modeling capabilities similar to BPMN. It is quite widely used as well.
I have a curiosity / thought sparked from this great article that is only adjacent to the topic at hand, it has nothing to do with Graph engineering and AI. I wonder why BPMN had some moderate amount of success in the 2010s, while other PML attempts failed. I was running the engineering team of a company called Appway, which built a BPMN tool for modeling and executing banking processes. We had a decent amount of success which lead to a great exit in 2021, and the competitive landscape was fierce. Companies like Appian, Pegasystems and the OS Flowable are still amongst the largest players in the space, alongside Mendix, Outsystems, Camunda and many more smaller companies.
The issues highlighted in the "what went wrong" chapter of this post are similar to what we faced when modeling banking processes with BPMN: the promise of the language was to lower the incidental complexity of coding them in C++/Java/C# and allow businesses to focus on the inherent complexity of the process at hand. The reality was that the formalism required to turn the BPMN model into an executable, required going down to the details and model each and every variable passing, logical condition and decision element. These complex models required the skills of SW Engineers to build and maintain, that at that point failed the promise of "getting the business closer to the application". However some of the commercial products I mentioned above are still quite successful in the enterprise landscape; I thought AI would deliver the final blow, but it seems like it didn't from the company performance of the aforementioned players (which aren't thriving, but they aren't failing either). What made the difference in BPMN / Workflow automation vs PML at large?
BPMN is simpler and more focused. Therefore, more effective. UML has some process modeling capabilities similar to BPMN. It is quite widely used as well.