Thoughts on Event-Driven Business Processes, Risk Mitigation, and Running a Real-Time Business
Meet the Experts
⇨ Event-based methods help firms to react effectively and rapidly to business events.
⇨ Machine learning can make these processes and methods even more efficient and effective.
⇨ Monitoring events in real-time give organizations the ability to alter key events.
Following the typical process, after checking the match between the invoice amount and the purchase order, an AP clerk calmly clicks the button to authorize payment. The payment advice to the bank is triggered and a large sum of cash is transferred to the vendor’s bank account. Unbeknownst to the AP clerk, however, the vendor’s payee address and bank account had recently been updated in the midst of concern over a possible banking crisis and a fraudulent bank account had been entered.
Crises happen, perhaps at an increasing pace, and the ability for firms to react effectively and rapidly to business events becomes ever more important. One of the novel ways in which this can be accomplished is to use event-based methods.
Such methods use real-time information on what’s going on in the business, and allow firms to develop responsive applications that work around the clock and that are always available to be triggered by specific business events. This new paradigm can be used to reimagine and build novel event-driven processes altogether, capturing the real-time, always-on nature of today’s commerce ecosystem.
We’ve seen methods grounded in events successfully used in several different processes, for example, to predict and react to changes in remaining cycle times, fulfillment timeline violations, and next process activities.
The opportunities that event-based architectures create are limitless, including the ability to reprioritize cases in real time to ensure adherence to service-level objectives, to predict process deviations, and to act on them in real time. Firms can also flexibly screen for compliance-related issues, such as the issue of bank account updates and subsequent payments, to stop cases that have a high probability of fraud, and put in place checks and balances for deeper review and approval.
An exciting feature of event-driven architectures is that this capability does not require changes to the core of the ERP system and can be deployed on an agile and quick timeline. This is particularly useful in the case of unforeseen environmental events that benefit from immediate system enhancement. Because the business systems emit the trigger events, any consuming application can dynamically attach itself to those events and carry out the required analysis and follow-on activity.
The event-driven approach can be extended into predictive event-based process monitoring by using information from prior iterations of business processes to create predictive models of how events should occur. These models are then compared to active cases running through the process in real time, measured by their emitted events and if there are any deviations from the accepted variances in process performance, exceptions are created and processed immediately.
Recent advances in machine learning, such as the development of long-short-term-memory neural network architectures, for example, have shown great promise in being able to learn and effectively predict complex event- and time-based processes. Further, developments in advanced Bayesian methods, which use prior probability distributions and dynamically update predictive models by including current data, show great promise in being able to model business processes and effectively detect process outliers in a robust way.
Imagine a world where there are no more struggles with managing delivery date expectations; particularly in make-to-order scenarios that take several third-party inputs for final product assembly, there are a multitude of variables impacting delivery dates. Once a customer order is received and confirmed, the delivery date is likely to change repeatedly. Historically, this process is managed with reporting, where a report of overdue inbound deliveries is issued and reviewed regularly, and action is taken on the outbound side, usually in a tedious, manual way.
In an event-enabled business process, this changes because the key events that relate to a customer order, such as input availability changes or carrier lead times, is monitored in real time by the system. When a change in one of these data points is detected, the system issues an event, which in turn triggers a functionality that updates the predicted delivery date for all related customer orders.
Now, the customer can be notified proactively, and even automatically, about a likely change in delivery date, so they can account for this change in their own plans. This is a case where reporting can potentially be fully replaced by an event-based mechanism.
Our team has recently seen a great increase in customer inquiries to discuss the architecture and benefits of event-driven business process, a wave encouraged by the capabilities within the S/4 platform to emit real-time business events to a central publish/subscribe system within its sibling Business Technology Platform. This enhancement of the platform allows it to dynamically trigger functionality based on specific business events, and to take the required action.
While event-driven business processes are still not mainstream adopted, I firmly believe that these process architectures show immense promise, particularly in their ability to improve the agility and resilience of firms, and by allowing rapid decoupled enhancements enabled by a unified event hub.