In any Process Automation scenario, the level of human intervention essentially defines how automated the process is. If there is no human intervention, then the process is by definition, fully automated. Some processes, for example in manufacturing, do not require any human decision making, and can be fully automated, without compromising on quality or efficiency. But in many cases, human choices are required at various stages within the process for the optimal result.
Business process management (BPM) software can coordinate the interaction between Human and Machine decisions within automated processes. Let’s examine six levels of decision-making within BPM processes:
- Straight-through process flow: These are stages of the process which are pre-programmed to follow a specific flow each time there are reached, without any deviation over process instances.
- Machine decisions based on business rules or logic: The flow of the process through these steps is calculated by the system based on a set of rules or business logic. The flow can vary each time, but the decision is entirely machine-based. The choice may be influenced by the initial input conditions (which may be human), or other environmental or system conditions, but there is no human intervention at the decision point.
- Machine decisions which made dynamically based on real time data/conditions: Such choices are made by the system at the point of the decision itself based on any number of factors including: environmental conditions, aggregated analytics, external system triggers. This level is similar to the previous level, though here, the choice cannot be predicted by the initial input conditions of the process.
- Machine learning scenarios: In this level, the system is pre-programmed to learn from past process instances, optimizing current choices based on the results of past choices. This can be viewed as a level of Artificial Intelligence, and is at the forefront of Intelligent Business Process Operations. Several BPM software vendors, such as AgilePoint, PNMsoft, and XMPro offer such machine learning functionality.
- Machine-assisted decisions: Here a human user is prompted to decide how the process will progress, but is assisted in this decision by the system, which can surface relevant data to help the user make an informed choice. This framework is optimal for situations which are unpredictable or dynamic, and cannot be foreseen using prior data. Also, it is necessary for cases where the system simply does not possess the human judgment or experience necessary to make the call.
- Human-only decisions: Here the process is fully controlled by the user, without any assistance. This scenario is ideal for cases where the system’s “opinion” or suggestion has no real value, perhaps in cases where the calculation is so complex or human-oriented that no assistance is valuable or necessary.
The end goal of all process automation is optimal performance. As each scenario is different, consider which level of process automation will help you best achieve that goal. Often, the answer is a mix of several levels, during different stages of the process.
Take for example the process of analyzing a silicon chip for defects in a fabrication plant. Most of this process is fully machine automated – once the technician places the silicon wafer on the analysis tool, the tool performs the defect analysis automatically, according to a pre-prescribed recipe of steps.
And yet, fabrication plants have discovered that this process can be optimized if at specific stages within the processes, the technician can be consulted for decision making. If, for example, a number of die sections have been found to be defective, the technician can be alerted and then make a machine-assisted decision on whether to abort or continue the analysis. The technician may possess more information or experience than the system which enables him to make a more intelligent choice (e.g. he is aware of a problem with the batch of wafers or the supplier), though he requires the system’s assistance in making this choice based on a complete picture of the data.
As such, even highly physical or chemical processes can be optimized if they are enhanced with human-decision points where appropriate. The governing principal should be that at any point, the choice should be made by the most capable participant – and sometimes, as a joint decision by multiple participants, each who are complementary to the other.