Robotic Process Automation (RPA) and AI

Overview of RPA

Organizations of all types and sizes can automate business processes by utilizing RPA software. Many RPA vendors offer automation solutions that do not require its users to have any programming skills; instead, they rely on users activating the robot to mimic keystrokes using a smart screen recording. Unsurprisingly, RPA adoption has grown substantially in recent years. Gartner notes that “robotic process automation has democratized integration and automation, leading to widespread business adoption” (Stoudt-Hansen et al., 2019). Adoption will continue to grow as organizations seek to realize productivity gains, increased accuracy, and cost savings through automation.

According to researchers van der Aalst et al. (2018), there are two characteristics of RPA that have contributed to its growth. First, RPA can layer on top of existing systems, enabling organizations to avoid redesigning or acquiring new platforms. Second, RPA software robots are flexible and can adapt to changes in the underlying information system. These two features of RPA imply that the investment can be relatively small compared to other IT projects. Consequently, RPA is currently viewed as a quick way to increase a firm’s return on investment (van der Aalst, Bichler, & Heinzl, 2018).

The automation of accounting processes has the potential to enhance the bottom line. By 2024, Gartner predicts that organizations can lower operational costs by 30% through redesigned operational processes and hyper-automation, combining RPA tools and machine learning applications (Stoudt-Hansen et al., 2019).

We defined RPA in Chapter 1 as a software application (commonly referred to as robots or bots) that automates a business process by replicating the actions of humans performing tasks, such as manipulating or transferring data within digital systems. RPA robots automate repetitive tasks, like macros in Excel, helping organizations get more work done in less time. However, a distinguishing feature of an RPA robot is the ability to interface with multiple applications and across functions. PwC describes RPA as “technology agnostic,” meaning that it can work across legacy ERPs, mainframes, desktop applications, and any other IT platform (PwC, 2017). In other words, unlike Excel macros, any technology platform that a human can use can also be navigated by an RPA robot.

Additionally, there is some debate over whether RPA should be considered a form of Al. A joint report by CPA Canada and the AICPA notes that RPA by itself does not constitute Al because it performs actions using pre-programmed instructions (CPA Canada & AICPA, 2019). The bots make no additional decisions, and no learning occurs after completing the tasks. Gartner seems to agree with this assessment, describing Al, ML, NLP as “not part of RPA per se, but are closely related” (Miers, Tornbohm, Kerremans, & Ray, 2019). We define Al broadly as a computer program or software application that can imitate or simulate human behavior. As such, RPA’s ability to mimic human actions places it on a spectrum where it could be considered an elementary form of AL UiPath, one of the leading RPA vendors, notes that “RPA robots utilize the user interface to capture data and manipulate applications just like humans do” (UiPath, n.d.).

Whether or not RPA is technically classified as Al is a matter of interpretation. What matters is that RPA is being used extensively in the accounting profession, often coupled with Al and ML technologies. Thus, accountants need to understand what RPA is and how it can be combined with Al.

RPA Vendors

There are currently many RPA solutions on the market. In a 2019 study, Gartner evaluated leading RPA providers based on the features of their solution, including integration via the user interface, large-scale data migration, and augmenting knowledge of workers (Miers et al., 2019). Examples of RPA solutions reviewed in the Gartner study included UiPath, Automation Anywhere, Blue Prism, Kofax, and WorkFusion.

When to Use RPA?

It is essential to consider when it is appropriate to use RPA before its implementation. According to a paper published in the Journal of

Emerging Technologies in Accounting, the implementation of RPA is appropriate when: (1) existing processes are well defined; (2) tasks are high volume and repeatable; and (3) tasks are mature, having predictable outcomes and known costs (Moffitt, Rozario, & Vasarhelyi, 2018). Gartner recommends that RPA automate business processes that are predictable and rules-based manual tasks that are well understood (Miers et al., 2019). Deloitte advises using RPA for operations that require manual intervention, are rules-based, and require a significant amount of time (Deloitte, 2017). These recommendations serve as a sound basis for accountants to use when trying to assess whether RPA will provide the desired benefits.

Gartner identified two deployment models for RPA platforms; (1) attended robots that augment human operators, and (2) unattended robots that have a separate and dedicated environment from all humans triggered by remote schedules or activity (e.g., through an API, or some observable event, such as the creation of a file in a directory or the arrival of an email into an account monitored by the robot (Miers et al., 2019).

Another study by Gartner focused primarily on the application of RPA in finance and accounting. It is recommended that RPA is applied for three use cases: (1) transferring data into or between AIS or ERP systems, (2) combining data into standardized reports, and (3) automating an existing structured business process or creating a business process (Tornbohm & Leiter, 2019).

Deloitte provides numerous examples such as transferring data, opening emails and attachments, connecting to system APIs, extracting data from structured documents such as invoices, logging into enterprise applications, performing calculations, collecting social media statistics, scraping data from the web, completing forms, and reading and writing to databases (Deloitte, n.d.).