Naïla and Serge, do you think Generative AI (GenAI) will have a significant impact on BPM?

Serge: When we look at classical, sequential business processes typically defined using BPMN (Business Process Model and Notation), GenAI might come into play with service tasks providing an input otherwise created by a human actor or supporting a human actor, thus increasing his productivity. This is typically what we already see with RPA (Robotic process automation) tools integrated into processes.

Naïla: You have to be aware that the effectiveness of GenAI depends on the quality of available data. It stands out as a valuable brainstorming aid, complementing but not supplanting human input. However, when faced with data beyond its training parameters, GenAI may generate misleading or inaccurate content, often indistinguishable due to its realistic nature. Ultimately, the combination of GenAI and human judgment can lead to more effective and efficient decision-making. Without human oversight or reliable data, the utility of GenAI diminishes significantly.

Serge: This leads to a second facet of BPM, business rules as described using DMN (Decision Model and Notation). Will we need DMN at all in the future? Why make the effort to precisely define business rules if GenAI can easily learn how to take decisions? Well, we often do not understand how and why GenAI comes to a conclusion. So if there is a way to deterministically calculate an outcome, this may be preferable. Especially if the consequences of an inaccurate decision are not acceptable. But what if it is too costly to define or execute the exact rule and the negative consequences of an inaccurate decision are not that relevant? Then, GenAI may be a good alternative. But let’s not forget that some decisions that we model are not fully deterministic. How would you model a doctor in an emergency room who within just a few seconds must decide on which treatment to apply to a patient in order to save his life? He will apply his expert skills, based on his experience.

Naïla: I would suggest not to forget that, just as humans, AI tools are not flawless. AI, when trained with data, can learn to make precise decisions in specific scenarios and act immediately. Unlike humans, it is not influenced by stress or emotions that might otherwise impact its decision-making. Therefore AI can be a great help, but it is not a panacea!

Serge: What about case management, the third approach in the BPM trilogy? Tasks are typically being activated by human actors based on their expert knowledge, correct? Yes and no. What I just described is the classical paradigm. But technically speaking, it is actually the sentries that allow tasks to activate based on IF and ON conditions. The if conditions are acted upon either by the human performer or by business rule tasks that set case file items. And now we are back to DMN, where now that we have widespread availability of GenAI, we can use this new technology side by side with deterministic algorithms. Suitability and economics will guide us when deciding which approach to take.

Naïla: Clearly, BPM is not only about sequencing repeatable steps, but it involves decision making and creativity. This is where we’ll see a lot of GenAI applications assisting human actors in the very near future!

Serge Schiltz is CEO and founder of processCentric GmbH and Naïla van Kommer conducts product work out in the areas of artificial intelligence and machine learning, she is also responsible for the visual identity of processCentric. 

You can read on the topic from other BPM experts by going to Zbigniew Misiak's full article on BPM Tips.