Traditionally, business processes have used the Internet of Things (IoT) as a source of data distribution. However, the current Business Process Model and Notation (BPMN 2.0) standard offers support for both modelers and IoT to generate low-code for IoT devices and execute aspects of business logic.
BPMN 2.0 is a standard created and maintained by the Object Management group that utilizes graphical notation for consistent business process communication. This article will focus on using IoT and machine learning with BPMN 2.0 to create workflows and the communication process between the execution engine and IoT devices.
Business processes can use IOT and ML for data-driven decision making
Many organizations have increased their usage of business processes to optimize their workflows. A business process is a collection of activities, tasks, resources, and stakeholders who collaborate, intending to manifest an outcome that produces value for the business and the customer. Areas such as supply chain, customer onboarding, account openings, and others can also gain a competitive advantage with IoT devices. How? Well, business processes can use IoT devices and machine learning for sensing, data capture, data storage, actuation, and to optimize the execution of data-driven decisions. In fact, IoT and machine learning can filter real-world data to make decisions locally where centralized management and coordination are not needed. However, since IoT devices are naturally heterogeneous, decentralization requires thoughtful design and execution.
BPMN 2.0 already offers modelers the option to define IoT, machine learning, and business processes using varying pools and specified via collaboration diagrams. Although, this needs to be connected via programming code executable by IoT and machine learning. Then, choose the right language, execute the code, and deploy it into IoT devices. The BPMN 2.0 performer class may then be used to define which IoT devices will take part in the business process. It’s crucial to have portable executable code used by varying IoT devices, even remote support, and programming.
IoT and machine learning use different forms of communication
Typically, web services provide IoT and machine learning the link to business processes with low-level details. Model languages already support web services and offer a centralized approach. Further, IoT functionalities are emitted as web services, which can be implemented through SOA tools or middleware. On the other hand, specific BPMN 2.0 extensions may include an IoT-specific process model.
For example, a graphical user interface (UI) can offer extensibility in designing IoT processes from a centralized portal while sending them directly as a sequence of operation calls to the intended IoT devices using proprietary communication protocols.
Make IoT and machine learning active participants in creating workflows
Invariably, IoT and machine learning can work in conjunction to manage and execute tasks within workflows, improving scalability, and reducing the number of necessary message exchanges. Nonetheless, using IoT and machine learning with BPMN 2.0 to create workflows requires a unified framework. BPMN 2.0 defines IoT behavior for targeting IoT device-specific code. Different code is generated based on the IoT device. Use the BPMN 2.0 resource class to add IoT device information.
Modeling IoT and machine learning behavior
BPMN defines behavior and interaction by using different pools along with collaboration diagrams. As a result, you can use standard BPMN 2.0 elements to model IoT and machine learning behavior. Consider the following elements:
- Flow control: Events, activities, and gateways
- Connecting objects: Sequence and message flow
- Data: Data objects
You can then use script tasks to read the IoT sensors and model the activation of the workflow. Moreover, the Performer class can define which resource performs the processes. And, the IoT device can be defined in three ways:
- Address
- Operations
- deviceType
Next, translate BPMN 2.0 processes into executable code that deploy and execute within IoT devices and machine learning. It helps to use code that patterns itself with well-established formal semantics and makes portable executable code for IoT and machine learning. You may also install the mobile code into IoT devices remotely. As such, start event messages can translate into function calls to trigger processes and workflows. Remember, IoT and machine learning are always ready to fulfill new requests.
Unquestionably, a successful proof of concept is the starting point. Usable IoT and machine learning can be facilitated through cloud services and in the architectural sense. Introducing IoT and machine learning into fully-manual workflows isn’t the most effortless transition. Use these questions to guide whether a workflow is ready for IoT and machine learning:
- How often is this workflow executed?
- How costly are the activities?
- How much runtime does it require?
- Would it make sense to automate the input and output of individual activities?
To illustrate, IoT and machine learning can learn from the history of manual classification of incoming business files and classify them similarly. A target key metric should be provided and optimized. A hit rate would describe the number of accurate vs. inaccurate classifications. The workflow improves when the hit rate goes up. As a result, you want IoT and machine learning to maximize the hit rate. Workflow design goals should be as follows:
- Simplicity (use a rule-based approach)
- Determinism
- Explainability (anomaly detection)
First, use a human-first decision support system where human decisions are required before IoT and machine learning is involved. Often, this type of support system can increase transparency for any misunderstanding, errors, or bias. IoT with machine learning can serve as quality control for any decision, whether it is low-level or high stakes, i.e., loan approvals, compliance, HR processes, and fraud detection. Then, IoT and machine learning can be trained consistently with new data when needed.
For example, a pandemic affected every business operation and workflow worldwide. However, when the COVID-19 pandemic becomes endemic, new processes and new data would be available to ensure relevance to current market conditions. Other scenarios, such as new medical device brands or car models, need collaboration between human-first decisions and machine learning.
Final thought
IoT and machine learning offer exciting opportunities to enhance BPMN 2.0 in creating workflows. Business processes can benefit from IoT devices’ sensing capabilities, networking features, and machine learning abilities. By translating BPMN 2.0 into portable code, you can successfully combine IoT, machine learning, and BPMN 2.0.
