A long-running decision is a special type of multistep decision that does not produce the outcome after execution instantly. Its execution will go to a dormant state. When the time is right, it comes back to life and continues the execution until it reaches another dormant state or the final decision is made, and it completes the decision cycle.
When a long-running decision is at the dormant state, it listens to interesting external events with specific topics, e.g. receiving an email, reaching a time/date, receiving a signal from a specific application, etc. Based on these external events, it will come back to life and continues the execution.
Stateful vs. Long-running Decisions
All long-running decisions are stateful, meaning they maintain their context and internal states integrity over time. This ensures they can continue the decision-making process from the last dormant point based on internal states and history of what’s happened already once they receive a particular event with a specific topic. However, not all stateful decisions are long-running. The stateful decision simply means it will manage states during execution i.e. it manages the states of business objects over time.
Long-running decisioning requires an advanced orchestration to coordinate execution and manage the internal states, routing tokens, and the current context of the decision.
A Case for Long-running Decision
There are some specific processes in any industry that are about making decisions. We obviously can look at them from the process lens, treat them as any other organization processes, and manage them using the same tooling: workflow management systems, BPM, custom code, etc., which they will suffer from the disconnected decisioning conditions.
The big difference between the decision-making process and other types of business processes is that they are more about decision-making than anything else. Decision-making requires more dynamic behavior and flexibility because they are influenced by rules, policies, regulations, and data. They need more strong data connectivity, processing, and validation. They will utilize different technologies for decision making i.e. data-driven decisions (ML) vs. rule-driven decisions.
Let’s have a look at a real example; in financial institutions, as part of the loan origination process, there is a sub-process of credit decisioning. This sub-process is regarding the creditworthiness of a particular applicant. In this sub-process, we can pull out the decision parts and let the loan origination process handles and coordinates everything. Or the better option is to look at the creditworthiness decision as a long-running decision.
This long-running decisioning model handles everything about the creditworthiness of an application. From receiving and validation and expiry of application to credit a custom decisioning algorithms and exception handling and sending notifications. It’s a high cohesion model which means it does what it is supposed to do and does it completely i.e. completes the full decision-cycle.
Last updated August 5th, 2021 at 01:50 pm, Published July 21st, 2021 at 01:50 pm