Robotic Process Automation (RPA) – Bibliometric Analysis and Literature Review

Technologies available within Industry 4.0 enable new opportunities for optimizing business processes in an organization. Currently used traditional methods of process improvement are insufficient to ensure the continuity and stability of processes in large organizations. The situation differs with companies that are at the initial stage of implementing the process approach because with the increase in process maturity, the demand for business process automation (BPA) rises.

Available publications cover the areas of knowledge dedicated to process management and organization maturity (Kosieradzka, 2022; Bitkowska, 2019; Dumas, 2017; Looy, 2014), methods of process mapping and simulation modeling (Briol, 2008; Rostek, 2020; Rother, 2003), methods of creative problem solving (Kosieradzka, 2013), and process classification (APQC, 2022, ISO 9001). Unfortunately, there is no systematic connection between the current state of knowledge and modern methods of BPA. Key source of expertise in this area is found in the practical knowledge of companies implementing specific automation solutions (EY, 2020; Deloitte, 2015; IRPA AI, 2017).

Process automation recalls the automation of production processes using, for example, welding robots and automation of production lines (Kaczmarek, 2020; Honczarenko, 2018). Business processes, such as financial, accounting, and transaction processes, are subject to a different type of automation. The goal of this automation is to replace repetitive, rule-based processes performed by a human with a bot. Examples of the transformation of business processes can be found in part II of the study (Sysko, 2021).

This article aims to identify the main research directions and trends in the scientific literature in the field of RPA. The author’s work goal is to build a transformation model for business processes in the field of BPA. It requires not only literature review in the context of general understanding of RPA definition, but also definition of what are the most common use-cases as well as if there are application models available. Based on BPA, the author narrows the research area to robotic process automation (RPA) as the most popular and widely used method.

BPA is about replacing human work with a computer program. It is a technology-based software to address rule-based processes. It supports the company’s operations to simplify business processes, improve service quality, and reduce costs, which is impacting revenue. Automation is a combination of application integration and resource restructuring (Skalfist, 2020). BPA goals are also as follows (Deloitte, 2022): simplification of business processes, increasing the share of automation in human work, cost reduction, increase in productivity, and growth of correctness in the implementation of processes.

Research shows that about 50% of jobs are susceptible to automation. This is especially true for professions that require simple skills (Nowak-Nowa, 2018).

Among the tools available for BPA, RPA is the best known. The implemented RPA solutions are met in 80% of the surveyed enterprises, and these companies indicate them as the first for conducting a case study of profitability (Delloitte, 2022). The extremely fast pace of growth of this part of the IT market can be confirmed by the results of research according to which the global value of the RPA tools market was estimated at USD 849 million in 2018, whereas in 2019, it was as much as USD 1.3 billion and in 2021, USD 2.9 billion (Delloitte, 2022). Also, in 2020, 73% of companies declared willingness to use the processes of intelligent automation (Delloitte, 2022). Therefore, the area of analysis was narrowed down to RPA method.

RPA is based on specific software that maps the processes. In RPA, the term “robot” means a license (Willcocks, 2015). Gartner proposes the following definition: “Robotic Process Automation is a tool that allows the user to configure one or more scripts (called “bots” by vendors) to automatically activate specific keystrokes imitating human work. As a result, bots can be used to mimic or emulate selected tasks (transaction steps) throughout the entire business or IT process.” RPA uses a combination of user interface interactions and descriptor technology. Scripts can overlap one or more applications (Gartner, 2022).

The term RPA was first introduced in O2 Telefonica by Lacity Mary and Willcocks Leslie (2016). In this article, we may find the approach of early adopters of RPA, and it gives a perspective for the future implementation. Actually, it can be said that the first RPA definition is mentioned there, and in further articles, we find citations to this one. This definition of RPA is remarkably similar to the one that is used currently: “For business processes, the term RPA most commonly refers to configuring the software to complete the tasks previously done by people. RPA software is ideally suited to replace humans who perform socalled ‘swivel chair’ processes. Such a process is where human sits in a swivel chair at a workstation and takes in work from many electronic inputs (like emails and spreadsheets), processes it using rules, adds data as necessary by accessing more systems and then inputs the completed work to yet another systems, like for instance ERP or customer relationship management (CRM) systems” (Lacity and Willcocks, 2016). The authors had a follow-up on the topic and described robots as a future of work (Lacity and Willcocks, 2016, Warwickshire) and as an innovative approach for automation services (Lacity and Willcocks, 2016; MITSloan).

In the article Sobczak (2022 p. 87–88), we can find definition of BPA – RPA enables the creation of software robots using graphic wizards or by recording human activities in such a way that:

it operates based on a given algorithm, but increasingly, they are enriched with certain elements of artificial intelligence (AI), thanks to which they can make complex decisions,

The article consists of four elements. The first one is “Methodology,” where description of the approach to systematic bibliometric review is given. The section “Results” highlights the outcomes of searching sources (Scopus, Web of Science, and IEEE Xplore databases) with statistical review, cluster analysis, and citations. The next one, “Discussion,” gives a follow-up on clusters definition as well as their review in most cited publications. The last one is “Conclusion,” where we can find the theoretical and practical meaning of the article for author’s study as well as further research steps.

As mentioned in the “Introduction” section, the main tool used in BPA is RPA. The research will focus on the application of process automation using the RPA tool. Again, emphasizing this area of analysis is crucial for the definition of the model for automation of business processes. The adopted methodology and approach are based on Szum (2021) and da Silva Costa, et al. (2022). 1)

Source selection, where Scopus, Web of Science, and IEEE Xplore databases were studied to extract keywords and the most cited publications. Analysis of the databases showed that these three chosen ones give the most appropriate results, as well as the number of searches gives the highest scores. Moreover, these databases have similar list of publications that support finding the most cited ones.

The steps of research method are shown in Figure 1.

Figure 1.

Application of the methodology described above allows to obtain outcomes given in Table 1. After implementation of steps 1–5, Scopus database gives the highest results, while Web of Science gives the lowest ones. The order changes while applying RPA shortcut to the analysis. This gives, in the author’s opinion, the best outcome concerning the topic of analysis. IEEE Xplore connects RPA to manufacturing processes, while Web of Science and Scopus give better results for BPA.

Moreover, if we take into consideration only publication year (Table 2 and Figure 2) for results from 2016 to 2022, we can see that starting from 2019, the topic RPA becomes increasingly recognizable. Ninety-three percent of the total number of publications from all sources was published after 2019, with the highest peak in 2022.

Figure 2.

Next step of the analysis is based on the type of documents. The results are presented in Table 3 and Figure 3. If we take Web of Science into consideration, more than 90% of results are proceeding papers (conference) and articles. In the citation research articles will be taken into consideration with amount of 67. If we take Scopus into account, more than 98% of results are conference papers, articles, and book chapters (Table 4 and Figure 4). IEEE Xplore analysis showed that more than 97% of publications were conference papers, standards, journals, and books. IEEE Xplore will not be taken into consideration in citation analysis due to dwindling number of articles available. This can indicate that the topic of RPA is well described in the conference papers and articles.

Figure 3.

General summary of the analysis of available publications results in confirming that the best sources for article review are Web of Science and Scopus.

Web of Science final results were also subject to graphical analysis by VOS viewer program. The graph is presented in Figure 4, which presents cooccurrence links between “Robotic Process Automation” phrase and other keywords mentioned in the 757 results, before adding the RPA shortcut to the narrowing criteria. The goal of this graph is to define connections to RPA based on the size of the circle.

What we can identify from Figure 4 is that RPA is often connected to robots used in manufacturing (left part of the graph). The right part of the graph gives an overview to RPA and connection to keywords like AI, business process, and so on.

The analysis of Figure 4 and article from Figure 3 provides the possibility to define clusters from the bibliometric analysis (Table 4). Four clusters were defined by the author with division to keywords included. These clusters will be used in further citation study. If citation analysis was taken into consideration, the analysis was limited to Web of Science and Scopus.

As the number of citations is one of the objective parameters, two more were added to the study – year of publication (2022) and topic of article connected to RPA implementation. As RPA is an emerging innovative technology, adding these criteria will give a more comprehensive view of the cluster analysis. The results are presented in Table 5. Ten publications in magazines were chosen as the most cited ones; four of them are case studies. The last four are strictly connected to the model of RPA implementation.

Bibliometric analysis shows that there are a limited number of publications available concerning the RPA topic. This is proven not only by the numbers, but also in VOSviewer for the Web of Science analysis – left part of the graph shows connection to manufacturing processes (i.e., manipulator, robotic arm). This shows that despite putting the best efforts into narrowing, still there are results that need to be eliminated from the analysis.

Moreover, in the topic of RPA, four clusters were identified during the research (Table 6), which are as follows: 1)

AI – This is a proof that RPA occurs in connection to this topic. Some authors claim that in the future, there will be a need to transform RPA solutions into more AI methods, which will make decisions, support solutions, and accelerate the learning process (Siderska, 2020). Sometimes, they are called cognitive automation.

Based on the list of citation analysis, it could be proposed to study them in the spectrum of RPA definition and cluster analysis. This approach leads to the result that all the articles have similar definition of RPA, which concerns bots, programs, software, and so on (Table 6). What is also important is the fact that cluster–business process appears in most of the sources. This is connected to the fact that RPA is presented as a case study of a specific process. Six publications provide some idea for method/model of RPA implementation, where basic steps are described and RPA potential is shown.

The bibliometric analysis provides compelling evidence that RPA is an emerging trend in the market. There are a limited number of publications available, but they allow to identify a strong relation between RPA and manufacturing robotics. The number of valuable publications rises in a trend of several per year. RPA is a new and not well-defined method yet. For the author’s further research, it is particularly important to find basic definitions and general conditions necessary to implement it in organizations. What cluster and publication analysis strictly also proved is the fact that the models available for RPA implementation give only a general idea of how to improve the organization. Most publications present RPA definition with a review and cost–benefits analysis. Of course, even though there are a large number of RPA vendors and products in the market, still there is an identified lack of implementation model guidelines, which this publication aims to address. Moreover, the available rules can be misleading and provide extremely limited number of case studies.

This confirms the thesis put in the beginning of the article that there is no direct connection between currently known methods and robotization. Where connections can be found it that RPA is used a method to digital transformation of companies.

The theoretical part of the study also proves that more in-depth analysis should be conducted in the context of RPA method and models of implementation to provide rules for processes that can be automated. At the beginning of every process robotization, a high-level analysis of RPA should be conducted to determine those processes with high RPA potential. Suitable conditions or rules should be strictly defined, so that they are easy to assess and implement. Moreover, the relation between process optimization and RPA should also be defined. In example should the process be optimized before automation or after. Next, the author would like to determine what organizational methods can support RPA model in the organization to set guidelines, which of them should be used, and when. Finally, the research should provide the steps of the RPA model in connection to business process management (BPM) lifecycle. This, in the author’s opinion, would result in a ready-to-use implementation model.

For further analysis, it is taken into consideration by the author that other methods of AI could be analyzed, such as robotic desktop automation (RDA). This method allows each business user to create a bot or voice bot applications that are currently widely used in customer relations.

After completion of financial and accounting processes analysis, in the author’s opinion, all processes that are classified as auxiliary/back office could be a subject to the RPA method. Moreover, in all kinds of enterprises that have a central hub of processes, the method can be applied.