Automation is the key to increase value of process discovery
How the complex nature of business processes and variations in human behavior make accurate process identification a difficult task and looked at automation as a solution
Add bookmark
Brief to process discovery
Before processes are automated, businesses need the necessary visibility to identify the most valuable ones for automation. For this purpose, many businesses have turned to process discovery, with research suggesting that the global process discovery and mining market will grow from US$670mn to US$4.3bn by 2025.
Process discovery is a technique through which every aspect of selected business processes is documented to create a visualization of how employees are performing them.
Its uses lie in locating bottlenecks and inefficiencies, as well as identifying the impact of variations within process steps.
The value of process discovery is dependent on how accurately a business can capture process-related data and apply it to form a visualization of the process flow. The challenge here is compounded by the complexity of business processes and variations in human behavior.
Drawing on insights from industry experts from the likes of UAE Exchange, Edrio and Nintex Kryon, this PEX Network report discusses why it is challenging to accurately identify businesses processes through manual process discovery and important to accurately do so. It looks into the value of automating the approach and shares expert advice along with a real-world example of successful process discovery.
“There are many unseen handshakes like touchpoints, data, tools, systems and documents that work closely together and create event logs when a process is carried out. These handshakes may become unidentified in the end-to-end process.”
Anna John
Senior PEX professional and transformation architect, UAE Exchange
The challenges of identifying processes accurately and efficiently
Complex business processes
One of the primary challenges practitioners faces when attempting to carry out process discovery initiatives is accurately identifying individual steps in organizational processes. One of the reasons that this is such a challenge is that many are complex by nature and contain many different touchpoints, actions and variations that make reliable identification difficult.
According to Rakhi Oswal, founder of online fashion start-up Edrio, the biggest challenge in manual process discovery involves inaccurate mapping of the business processes needed in organizations.
Anna John, senior process excellence (PEX) professional and transformation architect at financial services organization UAE Exchange, agrees and explains: “There are many unseen handshakes like touchpoints, data, tools, systems and documents that work closely together and create event logs when a process is carried out. These handshakes may become unidentified in the end-to-end process.”
Also read: Boosting customer experinece with attended automation
While process mining works well for processes that are already identified, it does not take into account process variations. This is where the value of process discovery lies when creating process maps. John notes: “Process models can flow in a linear or non-linear fashion and the process mining software, used to read event logs and identify and map the processes, fails sometimes due to the inability to capture the process outliers and variations.”
This can result in the inaccurate identification and measurement of business processes, which means that businesses attain a false view of how these processes function in reality.
Any opportunities that are uncovered for optimizing these processes may fail to achieve the desired results, potentially harming the business rather than helping it.
Another challenge is managing and connecting the data resources that are collected from the processes, as explains Oswal.
“Handling and connecting process data resources for bulk data transformation can slow the entire process,” she notes.
Variations in human behavior
Human behavior is a big factor that organizations need to take into account in process discovery initiatives. One may assume that if processes have defined steps for completion, they will always be carried out this way in practice.
The reality is that this is unlikely as any number of external factors can impact the way an employee completes a process, which creates an outlier. These include answering a phone call part way through the process or tiredness which can delay a process.
Daniel Peled, vice-president, global strategic sales for Nintex Kryon, notes: “Humans are not structured. We do not do processes step by step as we should, we start a process, go talk to a colleague, answer an email or a phone call and go back to completing it.
“In addition, once we know the process, we do it without thinking about the individual steps so any variations because of different data sets, information or scale are seen as the same for us but are actually different paths of doing the process.”
Given the complex nature of business processes and the difficulty of predicting human behavior, a solution is necessary for enhancing the accuracy of process identification. In the next section of this report, we explore how automation can do this.
The value of automating process discovery
The conventional procedure for identifying processes is manual and requires employees to document each individual stage of a task. This is time-consuming and, according to UAE Exchange’s John, has the potential to derail a project from the outset.
“Sometimes, this exercise alone can derail the whole project, because documenting each task, for example, opening a page, entering credentials, exchanging data, verifying criteria and moving data to different systems, is tedious for a human employee, especially when automating entire processes with robotic process automation and artificial intelligence,” she says.
Automating the collection of process data and the formation of event logs is the first step to enhancing the accuracy of process identification. John explains that the removal of human error, bias or fatigue that can cause inaccuracies in process identification will increase organizations’ confidence in the accuracy of their process event logs.
“Nowadays, there are myriad options to accelerate process identification and avoid the hassles of the manual approach by utilizing bots to identify the processes,” she says. “These automated methods will observe and record every interaction with various systems and create event logs for process modelling.”
An example of automated process discovery that utilizes artificial intelligence (AI) and machine learning (ML) algorithms to ensure the accuracy of process identification can be drawn from Wawanesa Insurance. The Canada-based insurance company was aware it needed more innovative solutions to manage and optimize its processes.
This was due to rapid changes in technology, the rise of InsurTech solutions and a demand for a better customer, broker and employee experience, especially in a post-Covid-19 world.
The company’s goal was to automate processes to reduce manual and repetitive tasks and help brokers and employees deliver a better customer experience by spending more time on value-adding tasks for customers.
To identify the best opportunities for automation, process identification was necessary and the business selected Nintex Kryon as a holistic automation, process discovery and analytical tool.
Through the tool the business was able to accurately identify and successfully automate 12 manual and repetitive processes. It helped save more than 15,000 full-time employee work hours and permanently improve the systems for data collection and entry. Aside from freeing up time for their employees to focus on value-adding tasks, such as customer inquiries and complaints, applying process discovery resulted in a significant cost saving of US$1mn for the business.
Peled recommends that organizations plan as a priority in any automated process discovery initiative, to achieve similar success to Wawanesa Insurance.
Also read: Overcoming major challenges in automation projects
“Define your goals, understand that process discovery is a part of a bigger theme enabling organizations to achieve process excellence and that by implementing process discovery you will have a lot more data about how your organization works,” he says. “Now the challenge is what you do with it. There needs to be a plan to take this data for process improvement and optimization, to identify the processes that can be automated and automate them. Planning is key to achieve success with process discovery.”
Automating process discovery accomplishes more than the simple offloading of manual tasks, however. The use of technologies such as AI and ML can take over the more analytical tasks that typically require human input, such as identifying where processes have deviated due to human behavior.
Leveraging AI and machine learning for process discovery
The use of algorithms in process discovery is a relatively new concept and became commercially available in 2018 when Kryon released its automated, AI-based process discovery technology.
The algorithms present in the process discovery platform apply AI and ML technology to automatically detect the tasks, applications, human interactions and variations that are associated with the completion of a process. They can also handle the analysis of the data and the generation of process models.
Nintex Kryon’s Peled explains: “[We can create] an algorithm that does not only find the repetitive steps we have done, but also connects the dots and sees the process the way a human has done it, for example understanding if an employee jumped on a phone call before continuing the process.
“The algorithms learn what variations we have in processes and captures them. [Since it was created by Kryon in 2018] we have continuously improved it and our process visualization feature, making AI-based process discovery much more accurate than having humans do it themselves.”
These algorithms will operate in the background of an employee’s desktop, for example automatically capturing every interaction and facet of the processes to understand how employees complete them. They will produce automatic analysis of the data captured and even infer what a user is doing to provide explanations for deviations.
These algorithms will constantly improve by learning from the process data they analyze. This will help them become increasingly adept at accurately modelling processes based on data collected and event logs generated, Peled explains.
“The only way to have an algorithm learn what a process looks like is via AI and ML. As we do more iterations of the process the algorithm better understands how it is structured in practice and defines it this way,” notes Peled.
“Once we know the process, we do it without thinking about the individual steps so any variations because of different data sets, information or scale are seen as the same for us but are actually different paths of doing the process.”
Daniel Peled
Vice-president, global strategic sales, Nintex Kryon
Key takeways of the report
With the popularity of automation on the rise, it is important for businesses to understand how they can best go about identifying those golden opportunities for automation through process discovery. It is equally as important to understand what challenges they can face when trying to accurately identify processes to overcome them and get process discovery initiatives on the right track.
This report has discussed how the complex nature of business processes and variations in human behavior make accurate process identification a difficult task and looked at automation as a solution. Organizations can remove human error and bias from task capture by offloading this task to bots and ensure that the data captured is truly representative of the business process.
Businesses can take this one step further and apply AI and ML to offload the analytical stages of process discovery and leverage these advanced technologies to generate true visibility over how business processes work.
