By Isi Osagie Ph.D and Tyler Safranek
The age-old challenge of IT departments has been to rid the perception of being a cost center to the business. The emergence of Machine Learning (ML) and Artificial Intelligence (AI) led process technology initiatives like Robotic Process Automation (RPA), Process Mining, and Intelligent Data Capture has put paid to that perception by guaranteeing operational efficiency and effectiveness. According to Gartner, the major reasons for the adoption of these technologies are to optimize operational efficiency, accelerate existing processes, and optimize cost.
Where To Start
Despite the quantifiable gains attributed to Intelligent Automation technologies, a new challenge persists. The plethora of technologies that make up IA can make adoption somewhat confusing. Questions like “which process deserves the most attention?”, and “which technology should be integrated first?”. Indeed, Gartner further suggests organizations should strive for an orchestrated, end-to-end, intelligent, event-driven form of automation, delivered with an effective combination of automation tools with multiple machine learning applications.
“The first rule of any technology used in a business is that automation applied to an efficient operation will magnify the efficiency. The second is that automation applied to an inefficient operation will magnify the inefficiency.” – Bill Gates
No organization sets out to magnify the inefficiencies of their processes. However, that is exactly what can happen if you are not able to make a detailed, objective assessment of your process. Methods like crowdsourcing or Subject Matter Expert (SME) led workshops for RPA process candidates are useful for generating ideas but can often plateau once the users’ requirements are met. Furthermore, these efforts are often subject to unconscious bias from SMEs and do not consistently provide objective criteria for several factors like RPA selection, Process Improvement, and Intelligent Data Capture or OCR led initiatives. To get a detailed view of your process and generate ideas for automation based on process data, we highly recommend starting your journey with Process Mining.
By transforming application log data from the systems where your processes are executed, Process Mining creates a “living” data model of your actual processes. Using this process model, you can then measure cycle times, deviation rates, and automation potential over your entire process rather than taking a small sample of cases and extrapolating the results. This objective, live view of your process allows you to simultaneously make better decisions about which opportunities to focus on as well as monitor improvement and automation changes to your process in real-time.
Fig. 1 The Process Graph above displays the flow of activities in the Accounts Payable Process. Activities highlighted in blue represent activities that are sometimes automated; the darker they are the more frequently they are performed by an RPA bot.
To illustrate how this works, let’s look at an example from an Accounts Payable process. As can be seen in the diagram above, the happy path for the process contains five steps; however, using process mining we also identify several process steps and activities that highlight the inefficiencies in the process.
Data transmission and errors in the send/receive process leads to human intervention and often rekeying of invoices which is both time consuming, repetitive, and subject to error. The question is: how often is this occurring in your process at a detailed level?
Looking at the process graph, we can see that the “Check Received Invoice” activity is never automated, as indicated by its gray color in the graph. We also see that we “Request Data” from our vendors over 5,000 times. All these manual interventions in your process decreases your workforce’s productivity and adds to the overall cycle time of your process. Now that we have identified a process gap, we can evaluate the best solutions for closing it.
In this example, process mining was used to quickly (and accurately) identify the following:
- Activities where data transmission was interrupted,
- Activities that are most suitable for automation and,
- Activities that would benefit from machine learning led IDC technologies.
Armed with the value assessment and detailed information about which areas to focus on, you can then determine what is the right solution and which areas to target first.
Intelligent Data Capture
Our Intelligent Data Capture (IDC) solution builds on from previous template-based transformation initiatives to what we have now in ML-based OCR integrated solutions that are faster, more efficient, and as effective. A system that does not require vendor-specific file systems; but utilizes machine learning and analytics to build on general fields of information to analyze thousands of documents effectively and efficiently. In the example above, the first five steps in the happy path of the process would significantly benefit from integrating IDC and RPA. Using our IDC platforms, the “Receive Invoice” and “Check Received Invoice” would be recognized and validated, the output from this would then be further validated by an RPA bot against the organization’s database. Here, all errors are immediately identified, and an email would be sent to the appropriate party to correct the discrepancy.
In summation, each individual technology described here provides a positive impact on your process but an orchestrated end-to-end utilization of all three intelligent automation platforms will significantly reduce guesswork, and enhance predictability and enable better decision making and increases the ROI from any RPA transformation project.
To get started on your process mining journey, lookout for our webinar and subsequent blogs on adopting process mining.