In a time of remote work, automation has quickly risen in importance to become a top priority for finance and accounting departments. As we have learned to conduct foundational business activities on a fully remote basis, companies need to standardize and accelerate core processes. In this light, the phrase “Robotic Process Automation” (RPA) comes up in thought leader discussions and search results as a panacea to solve this problem. The value of RPA has been documented by a variety of industry analyst and consulting groups and Amalgam Insights estimates that RPA can typically result in 60-70% savings in the processes that are supported.
But behind the hype, the phrase “Robotic Process Automation” is a bit confusing to those who aren’t deeply immersed in the topic, which is understandable since the term is relatively new. Prior to 2016, this phrase was searched on par with the phrase “business process automation,” the historic term for mapping and deploying process automation.
And though Robotic Process Automation has taken off rapidly as a term, the phrase itself is still somewhat opaque. We all understand the concept of process automation, but what exactly makes RPA “robotic”? These tasks typically transfer data from one electronic source to another and may involve some sort of calculation to contextualize the data. For instance, data that needs to be transferred from one source to another based on specific formatting or governance requirements could be seen as a “robotic” task that would theoretically be better performed by an automated set of rules rather than by the manual efforts of human auditors and data analytics.
From a technology perspective, a process can be seen as “robotic” if it is repetitive, routine, rules-based, and recurring. To elaborate:
Repetitive: Processes that need to be conducted over and over again. Think about copying and pasting back and forth, switching between two or more apps, comparing two or more columns for formatting and data quality inconsistencies. These are processes that are typically seen as mind-numbing to conduct at scale, especially when they are done thousands of times per day.
Routine: Good RPA candidate processes are tasks that employees currently support as part of their day-to-day job. From a practical perspective, this may include transaction reconciliation and data cleansing tasks conducted every month as part of a financial close.
Rules-Based: RPA tasks need to have well-defined rules, which can provide their own challenges of process mapping and documentation. RPA is poorly designed to handle exceptions and can quickly get overwhelmed by errors due to the speed that it handles transactions. Robotic automation is fairly limited in terms of making judgment calls or recommendations. RPA should be seen as a way of offloading high volume, data-intensive, and rules-based tasks away from humans both to increase the speed of processing and to provide a consistent digital approach for supporting transactions.
Recurring: RPA tasks should also be activities that need to be conducted on a fairly regular basis, either because they need to happen on a monthly or quarterly basis or because they are associated with common business activities.
To contextualize these qualities, a few key opportunities for RPA in finance and accounting departments include basic data format validation, near-real-time data updates, supplier and customer onboarding, pricing and discounting checks to support complex quotes and purchases, transaction reconciliation for financial close, governed filings, and journal reconciliation.
Combining AI and RPA for Enhanced Results
As companies consider RPA as an enabler to support increased productivity, they need to combine the large data and analytic processing requirements with emerging machine learning and artificial intelligence capabilities to take full advantage.
From a tactical perspective, artificial intelligence can provide logic to further accelerate RPA efforts by finding the best data for process workflows. This may be as simple as finding better ways to identify date formats or using a combination of metadata, AI-developed ontologies and taxonomies, and third-party data to identify more accurate business drivers and data that should be used for a specific process automation workflow.
In addition, although the matching, formatting, checks, and validation associated with RPA are relatively basic in nature, the sheer volume of transactions supported in RPA can result in unexpected challenges in processing the full variety of scenarios that may show up. Artificial intelligence can help conduct the forensics of process mapping and rules definitions used to document processes. This helps companies to accelerate the time it takes to set up RPA for specific tasks. Amalgam Insights believes this is an opportunity both for companies to invest in their own data scientists to accelerate data-intensive and repetitive processes as well as to work with vendors that focus on using AI to accelerate specific processes.
AI also provides opportunities to model the data created, updated, and transformed in RPA tools to support more strategic insights. For instance, there may be specific transactional trends that are increasingly out of scope or out of compliance. With manual processing, these trends may simply be corrected or treated as “needle in a haystack” occurrences, but modeling and data lineage at scale could potentially discover upstream process issues that have additional repercussions for the company at-large.
Conclusion
AI and RPA are two of the trendiest technology trends in enterprise software. Even in a quarantined work environment, both of these terms are seen as potential solutions to improve businesses. However, businesses pursuing automation efforts need to have realistic expectations both of what to expect from RPA and what is needed to enhance RPA from a basic operational upgrade to a strategic business enabler. In this light, Amalgam Insights provides the following guidance.
First, if you’re thinking about using RPA to both standardize and accelerate boring and repetitive work within your finance team, focus on documenting rules and exceptions. It is typically fairly easy to describe the ideal way that a process should work, but the challenge tends to be in effectively identifying and solving exceptions with a rule-based method. This may be an opportunity to augment manual process audits and process mapping by using machine learning to document processes based on tracking the entire lineage of a process from data source to business outcome. This may also be an opportunity to provide a reality check to the business seeking to automate processes that should be better left as human-monitored processes, at least until RPA handles exceptions more intelligently.
Second, look for both generalized RPA solutions that can be used to support generic data cleansing and processing tasks as well as use-specific RPA built to handle specific departmental or vertical use cases, such as finance. Because RPA is a rules-based technology, it ends up being only as useful as the set of business rules that are documented and encoded in the solution. As processes become more complex and more dependent on data sources that are susceptible to change, it becomes increasingly important for the RPA to be managed by a subject matter expert that can quickly document, parse, and set up new data checks, transformations, and transfers.
Third, RPA should not simply be seen as an opportunity to “do more with less.” Although automation is an opportunity to reduce the assets and resources associated with supporting rules-based analysis, the biggest opportunities for RPA occur when they are used to provide people with the information needed to identify trends, insights, and the opportunity for process changes. RPA is a strong tool for accelerating the throughput of processes, but typically lacks the business context and understanding to reconcile outputs in a meaningful way. This means that from a practical perspective, RPA rules need to be extremely well defined or they will end up pushing exceptions that may not have any actual business issues back to human reviewers. And then, RPA needs to be paired with both analytics and AI capabilities to fully understand trends and anomalies within the activities being automated.