VUCA and Business Process Management (BPM): an Introduction — Part III

Brief presentation of some interesting BPM automation solutions: process mining, robotic process automation, low-/no-code ülatforms, artificial intelligence and hyperautomation

Table of Contents

Interesting BPM automation solutions
1. Process Mining
2. Robotic Process Automation (RPA)
3. No Code/Low Code Platforms
4. Artificial Intelligence/Machine Learning
5. Hyperautomation
Key Takeaways/Tip

In the previous parts of this introduction to Business Process Management (BPM), we discussed some fundamentals (part I), including an iterative lifecycle consisting of 7 phases (part II).

In this third and final part, we present some interesting business process automation solutions that can be combined with the process/workflow engines traditionally used to execute business processes containing the appropriate technical details.

Interesting BPM automation solutions

1. Process Mining

We briefly introduced process mining in part II of this BPM series when we talked about automatically deriving business process models from event logs. In this section, we dive a little deeper into this exciting topic.

In general, process mining can be seen as a bridge between the scientific disciplines of process science and data science. In particular, the sub-discipline of data mining, in which large amounts of data (big data) are searched for interesting patterns and processed for users.
Or to put it another way, process mining closes a gap between those two disciplines according to spiritus rector Wil M.P. van der Aalst (“Process Mining: Data Science in Action”).
This is the case because, on the one hand process science tends to develop models without considering the associated (real-time) data in detail. On the other hand, data science doesn’t deal with (business) processes.

More specifically, process mining involves analyzing and improving processes using data contained in event logs, which have at least the following three mandatory attributes of an event:

• A case ID attribute that uniquely identifies a specific instance of a process.
• An activity attribute that describes a task being performed.
• A timestamp attribute that specifies the exact time at which the task is being executed.

An event is then a row in an event log where these three attributes can be used by process mining tools to determine and verify the control flow of a process (Note: There may be additional event attributes referring to resources, costs, etc. but these are optional).

To illustrate what has been said, the following is a brief excerpt from an event log related to making pizzas. In addition to the three mandatory attributes in this example case ID, activity and timestamp the two optional attributes resource and customer are used:

Pizza production event log (W.M.P. van der Aalst, “Process Mining: A 360 Degree Overview“, in: W.M.P. van der Aalst/J. Carmona (eds.), 2022, “Process Mining Handbook”, p. 13)

Some of the IT systems that generate large amounts of such event data in organizations are for example the following:

• Workflow Management Systems
• Business Process Management Systems (BPMS)
• Enterprise Resource Planning (ERP) Systems
• Customer Relationship Management (CRM) Systems

According to the Process Mining Handbook (pp. 23–28), these event data can then be used by six types of process mining techniques:

• Process discovery:
  - The main goal is in this context to create a process model from the event log.
  - This technique is often used when documentation of the process cannot be obtained through other (manual) approaches or when the quality of existing documentation is questionable.
• Conformance checking:
  - Here, the behavior recorded in event logs and an existing process model are compared.
  - This technique is useful in business alignment or for auditing processes within an organization.
• Performance analysis: Improving processes by discovering problems is a common goal of process mining. In addition to possible conformance issues, these may also be performance issues that need to be addressed.
  Performance problems in this context could mean:
  - Failure to complete an operation on time.
  - Limited production.
  - Missed deadlines.
  - Lateness.
  - Excessive rework and recurring quality problems.
• Comparative Process Mining: This mining technique compares multiple event logs to answer the following questions:
  - What are the noticeable differences and similarities between these logs?
  - What factors lead to these differences?
  The observed differences can then be explained by root cause analysis.
  In addition, comparative process mining is well suited for benchmarking between or within organizations.
• Predictive Process Mining: The previous four process mining techniques are backward-looking. However, it’s also possible to use the process models discovered and enriched by process mining in a forward-looking way.
  In this case, standard machine learning techniques (regression, decision trees, etc.) can be applied to models that can then be used in a predictive way.
• Action-Oriented Process Mining:
  - Predictive process mining can lead to diagnostics of the current process state. These diagnoses can be transformed into improvement actions, e.g. using low-code automation platforms (see below) as part of this final process mining technique.
  - Process mining can also uncover repetitive and tedious work that can be automated using “software robots” without changing the underlying IT systems. These software robots are also known as Robotic Process Automation (RPA), which leads to the next topic.

Before we move on to RPA, here’s an overview of the six process mining techniques discussed above:

Six frequently used types of process mining (W.M.P. van der Aalst, “Process Mining: A 360 Degree Overview“, in: W.M.P. van der Aalst/J. Carmona (eds.), 2022, “Process Mining Handbook”, p. 24)

2. Robotic Process Automation (RPA)

Robotic Process Automation (RPA) is another BPM automation trend that has become increasingly popular in recent years.

RPA refers to the use of “software robots” (or “bots” for short) to automate specific tasks that have the following characteristics:

• They are repetitive and tedious.
• They have a well-defined set of rules and structured data.
• They have a high volume of transactions.
• They are typically part of larger business processes.
• They occur at the user interface (G)UI level.
• They involve activities such as: data entry or extraction, form filling or simple decision making.

The RPA bots then automate the user’s interactions with the (G)UI, similar to GUI testing tools, where the demonstration actions performed by the user are recorded and replayed.
However, a key difference between RPA and GUI testing tools is that RPA enables the processing of data within and between multiple applications.

Compared to traditional workflow automation solutions that require a developer to know a business process inside out, using RPA has several advantages:

• Non-invasiveness: RPA is non-invasive because it works at the (G)UI level.
• System integration: It can integrate with existing IT/legacy systems that lack application programming interfaces (APIs) or require manual intervention.
• Ease of use: RPA is easy to use because its “record and play” technique doesn’t require complex programming and quality testing like traditional automation (Note: An RPA bot should still be tested on a smaller scale before scaling it up for full deployment!)
• Speed of implementation: RPA solutions are faster to implement than traditional automation solutions.

However, there are also several challenges to using RPA. For example:

• Process variability: Processes with high levels of variability or exceptions can be a challenge for RPA implementation. That is to say The extra “bot layer” can easily break!
  Handling such variability may require more complex rule structures or additional decision-making capabilities.
• False expectations about cost and support due to the RPA hype: The cost and ongoing effort of many RPA implementations tend to end up being higher than originally anticipated.
• Security and compliance: RPA solutions should be designed with appropriate access controls and privacy measures to ensure that the bot follows security protocols and meets compliance requirements.
• Change Management: Adopting RPA can lead to changes in work practices and employee resistance.
  Therefore, top and change managers should make it clear that RPA is not intended to replace humans, but to augment them (!) by freeing them from repetitive tasks that are both time-consuming and tedious.

In short, don’t believe the RPA hype and use it as just another tool in your BPM toolbox!

3. No-Code/Low-Code Platforms

The third solution for automating business processes is low-code or even no-code platforms. Both of these approaches are gaining popularity because they democratize coding.
This means that users with limited or no programming skills — so-called “citizen developers” — can now quickly build applications and automate processes.

Isn’t low-/no-code the same as RPA in the domain of BPM?
The answer is no. Low-/no-code approaches and RPA are not the same thing, although they both contribute to process automation.

Here’s how they differ:

• Low-code or no-code platforms provide visual interfaces, drag-and-drop functionality, and pre-built components that simplify application development.
  Users can then configure and customize processes, workflows, and user interfaces using a visual modeling approach.
  In other words, low-/no-code platforms focus on empowering business users to quickly build applications without extensive coding.
• RPA, on the other hand, focuses on automating specific tasks at the (G)UI level. This is done using the “record & replay” technique mentioned above within existing applications.

In brief, both of these automation solutions have their raison d’être in BPM: Low-/no-code platforms provide a broader development environment for building applications and automating end-to-end processes, while RPA targets the automation of specific repetitive tasks that require human-like interaction with software systems.

In addition, low-/no code and RPA can complement each other, enabling a more comprehensive automation solution that leverages the strengths of both approaches!

As was the case with RPA, you would be well advised not to blindly trust the hype of the “citizen developers” and to use low-/no-code as an additional tool in your BPM toolbox!

For an overview of the pros and cons of low- and no-code platforms, see also these introductory articles written in the context of Microsoft’s Power Apps:

• Low-code vs. no-code app development
• What is a low-code development platform?

4. Artificial Intelligence/Machine Learning

While generative AIs like ChatGPT, Bard, etc. are taking the world by storm, Artificial Intelligence and especially Machine Learning (ML) techniques are also being used in business process automation to increase efficiency and accuracy.

The procedure is as follows:

• First, identify the business processes that are suitable for automation using AI techniques. These are usually processes that involve repetitive tasks, decision-making, data analysis, or pattern recognition.
• Second, collect relevant data that will be used to train the machine learning models. This may include the data sources such as:
  - Historical process data.
  - Transaction logs.
  - Customer interactions.
  - Or any other data source that provides insight into the process.
• Third, identify and extract meaningful features from the collected data that are relevant to the process you want to automate. This involves the transformation of raw data into a format suitable for input into machine learning models.
• Fourth, select the appropriate machine learning algorithm or technique, i.e. (supervised, unsupervised, etc.) learning, based on the nature of the process and the data available.
• Fifth, train the selected machine learning models with the collected data. This involves feeding the data into the models and iteratively adjusting the parameters of the model to improve its performance.
• Sixth, integrate trained machine learning models with workflow systems. This may involve the creation of APIs or interfaces to enable data exchange between the models and other systems.
• Seventh, configure the automated system to use deployed machine learning models to make real-time decisions. To guide the automated process, the models can analyze incoming data, predict outcomes, or make recommendations.

In a nutshell, AI/ML techniques can be applied to automate cognitive tasks such as predictive analytics, pattern recognition, etc. which can also be used in business automation.

5. Hyperautomation

Hyperautomation as a convergence of traditional workflow automation using workflow engines/business process management systems (BPMS) and the automation solutions described above is a kind of “next frontier” in business (process) automation.
The goal is not only to automate more processes (e.g. those that are undocumented and use unstructured data) but also to be more impactful (“intelligent”) than traditional automation capabilities. This means that hyperautomation leverages advanced technologies to automate both routine and complex tasks at an unprecedented level. This can:

• Increase efficiency.
• Increase agility/flexibility.
• Improve decision making and customer experience.

IBM goes even so far as to claim that:

Hyperautomation is the concept of automating everything in an organization that can be automated.

Of course, it remains to be seen what will remain of this claim once the: hype sow has been driven through the BPM village and some calm has returned.
In the meantime, here are some examples of what hyperautomation projects can look like: Calkins, M. et al. (2020), “Hyperautomation”.

Key Takeaways/Tip

Here are the key takeaways from Part III of this introduction to BPM, which deals with the solutions/trends in BPM automation:

• Process mining builds a bridge between process science, which tends to develop models without considering the associated data in detail, and data science, which tends to ignore (business) processes.
  We can then distinguish the following six types of process mining techniques that all rely on the large amount of event data produced by IT systems such as workflow engines, business process management systems, etc.:
  - Process discovery
  - Conformance checking
  - Performance analysis
  - Comparative Process Mining
  - Predictive Process Mining
  - Action-Oriented Process Mining
• In Robotic Process Automation (RPA), the corresponding “software bots” use a “record & replay” technique to automate especially the user’s interactions with the (G)UI within and between multiple applications.
• Low-/no-code platforms provide a development environment that enables “citizen developers” with no or limited coding skills to quickly build applications and automate end-to-end processes.
  This distinguishes the low-/no-code approach from RPA, which is about automating specific repetitive tasks that require human-like interaction with software systems.
  However, it‘s’ also important to note here that the low-/no-code and RPA approaches can complement each other and aren’t mutually exclusive!
• Artificial Intelligence (AI)/Machine learning (ML) techniques are also useful in business automation through the automation of cognitive tasks such as predictive analysis, pattern recognition, etc.
• Finally, hyperautomation is the latest trend where the various automation solutions described above (including traditional process/workflow management systems) converge to take business automation to unprecedented (“more intelligent”) levels.

Final tip:
Introducing business process automation often requires changes in work practices, roles, and responsibilities. Consequently, the likely resistance to change and concerns about job security are challenges that must be addressed from the outset (!) through effective change management strategies and practices.

Addressing these challenges requires:

• The support of top management.
• Effective communication with all stakeholders.
• A team of enthusiastic multipliers who can make a difference in the organization.
• Strong analytical capabilities.
• Data governance practices.
• Collaboration between business and IT teams.

If this doesn’t happen, it’s likely that the culture of the specific organization (understood here as a complex social system whose countless communication streams are ultimately impossible to control) will eat the business automation project for breakfast — to paraphrase a famous quote by Peter Drucker regarding strategy.

In brief, the road to the already crowded graveyard of failed digitization projects is paved with non existing or bad change management strategies/practices.

Thanks for reading and, hopefully, see you in the next post Why organizational / process culture? Part I: background questions!

Author for WAITS Software und Prozessberatungsgesellschaft mbH, Cologne, Germany: Peter Bormann — July 2023.