RPA requires coding

Selection of business processes for the application of Robotic Process Automation using the example of an insurance company


Frequently, low IT resources, missing software interfaces or an outdated and complex system landscape slow down the automation of business processes. Robotic Process Automation (RPA) is a promising method to automate business processes on a surface-based basis and without major system interventions and to reduce media breaks. The selection of the right processes is crucial for the success of RPA projects. The present article provides selection criteria for this, which result from a qualitative content analysis of eleven interviews with RPA experts from the insurance sector. The result includes a weighted list of seven dimensions and 51 process criteria that favor automation with software robots or whose non-fulfillment makes implementation difficult or even impossible. The three most important criteria for selecting business processes for automation using RPA include the relief of the employees involved in the process (employee overload), the feasibility of the process using rules (rule-based process control) and a positive cost-benefit comparison. Practitioners can use these criteria to make a systematic selection of RPA-relevant processes. From a scientific perspective, the results provide a basis for explaining the success and failure of RPA projects.


Often, low IT resources, missing APIs or an outdated and complex system landscape slow down the automation of business processes. Robotic Process Automation (RPA) is a promising method to automate business processes in a UI-based manner and without major system changes as well as to reduce media breaks. The selection of suitable processes is crucial for the success of RPA projects. This article provides process selection criteria, which result from a qualitative content analysis of eleven interviews with RPA experts from the insurance sector. The result is a weighted list of seven dimensions and 51 process criteria that favor automation with software robots or whose non-fulfillment makes implementation difficult or even impossible. The three most important criteria for selecting business processes for automation using RPA include the work load reducton of the employees involved in the process (employee overload), the executability of the process using rules (rule-based process control) and a positive cost benefit comparison. Practitioners can use these criteria to make a systematic selection of RPA-relevant processes. From a scientific perspective, the results provide a basis for explaining the success and failure of RPA projects.

Introduction and motivation

Frequently, low IT resources, missing software interfaces or an outdated and complex system landscape slow down the automation of the execution of business processes (Syed et al. 2020; Beetz and Riedl 2019). Robotic Process Automation (RPA) is a promising method to automate business processes on a surface-based basis and to reduce media breaks. In contrast to traditional forms of system-based process improvement, software robots do not require any invasive interventions in the system landscape. User-friendly tools such as recorders and “drag-and-drop” modules enable experts without in-depth technical skills to train robots and automate their own processes.

The number and complexity of the existing business processes require a systematic approach in order to identify the processes for which RPA can be used profitably under the given framework conditions. When selecting the right processes, the properties of the business processes play a special role (van der Aalst et al. 2018). Studies have already been carried out on relevant properties and initial suggestions for selection criteria developed (e.g. Syed et al. 2020; Beetz and Riedl 2019; Lacity and Willcocks 2016; Fung 2014). Qualitative empirical studies have so far been the exception.

The present article addresses the practical survey of selection criteria and, based on an interview, answers the question of which properties a business process should have in order to be profitably supported by RPA. As a result, this article provides a weighted list of process criteria that favor automation with software robots or whose non-fulfillment can make implementation difficult or even prevent it. The contribution ends with suggestions for further research based on the results.

Robotic Process Automation and selection criteria for processes

Robotic Process Automation (RPA) describes an emerging form of business process automation (Aguirre and Rodriguez 2017) that uses virtual robots to operate a company's existing application systems (Sutherland 2013). Contrary to the idea of ​​physical robots operating in an office, it is software that is given the name “robot” because it imitates human interactions on user interfaces (Lacity and Willcocks 2016).

Unlike conventional process automation tools, RPA is a non-invasive technology that does not require any changes to a company's existing IT infrastructure (Beetz and Riedl 2019; Willcocks et al. 2015). Rather, the technology acts in an "outside-in" manner on the existing system landscape by only accessing existing system interfaces such as user interfaces (Sutherland 2013; van der Aalst et al. 2018).

For this purpose, a robot can be "trained" or configured by a specialist without IT knowledge. No or only limited programming knowledge is required (Willcocks et al. 2015). In contrast to most BPMN software packages, no programming skills are required to configure the robots (Aguirre and Rodriguez 2017).

Initial work has already been done on the selection of suitable processes (e.g. Syed et al. 2020; Beetz and Riedl 2019; Lacity and Willcocks 2016; Fung 2014). In particular, the number of process instances and systems involved as well as rule-based process control are mentioned again and again. What existing contributions have in common is that they mainly use argumentative and experience-based selection criteria. Expert interviews such as those in Beetz and Riedl (2019) and Fung (2014) have rarely been carried out, which motivates this article.

methodic procedure

In order to obtain criteria relevant to practice for the selection of business processes for RPA, a qualitative content analysis with inductive category formation according to Mayring (2010) was carried out. In order to achieve a high quality of results, only people who had experience in at least one of the methods proposed by Smeets et al. (2019) have collected defined roles of an RPA unit (RPA manager, RPA business analyst, RPA developer and RPA solution architect).

To carry out the interview, a guide was developed, the structure of which is divided into five phases. After a short idea, a thematic introduction and the Warm-up phase follow in Bulk Specific questions about the benefits of RPA for the company, the comparison of suitable processes or their properties and the most important selection and exclusion criteria from the point of view of the expert. In addition, important sources of information and influencing factors should be determined and a comparison should be made with other automation concepts or manual processing. In the Cool-off phase the interview partner is given space for additional explanations. The interview ends with a adoption as well as information on the anonymous evaluation and the next steps. All interviews are recorded and transcribed using a digital recording device.

The analysis includes the rule-based assignment of the text material to inductively derived process criteria. Based on Vogelsang et al. (2013) the development of the criteria, the determination of their clear definitions and the coding of the material form an iterative process that is also used in the analysis carried out. To the relevance To evaluate the statements, the frequency of the category mentions is collected and compared. For this purpose, the coded statements are made using the adapted procedure from Vogelsang et al. (2013), grouped into negative (−1), neutral (0), positive (1) or particularly positive (2) statements.

The sum of the relevance values ​​of a category serves as a key figure for sorting the selection criteria. In addition, the mean value of relevance forms a threshold for above-average relevant criteria. Another key figure is the Interview frequency which reflects the number of experts who commented on a particular category (Vogelsang et al. 2013). The basic assumption is that categories with a high interview frequency are more important (Vogelsang et al. 2013). In addition, the entirety of all coded statements of a category is represented by the frequency expressed.


Investigation scenario and overview of the selection criteria

The investigation was carried out at a medium-sized German insurance group, which carries out a three-digit number of business processes, and a consulting company working for the insurer. At the time of the investigation, the group is in a phase of rationalization that is to be accompanied by the use of IT, among other things. Both individual subsidiaries (TG) and the parent company (DG) of the group as well as a consulting company as an external service provider (ED) are involved in the investigation. A total of eleven RPA experts were interviewed, with their technical expertise at least one of the RPA roles according to Smeets et al. (2019) must correspond. Tab. 1 contains the demographic data of all interview participants.

The transcripts of the eleven expert interviews result in 576 statements, which are summarized in a total of 51 inductively formed categories or selection criteria. Based on the selection criteria, the dimensions process stakeholder, qualitative process properties, quantitative process properties, risk and compliance, information systems, task areas and development are formed and the individual criteria are assigned to these dimensions. The interview frequency, the frequency of the allocation of coded text passages (statement frequency) and the relevance values ​​are shown in Tab. 2. The mean of the relevance values ​​is ten. Grayed-out selection criteria are below this limit value and are not explained further in this article. The expert ID from Table 2 is used to refer to statements made by the interview participants.

Process stakeholders

The process stakeholder dimension depicts criteria that are related to those involved who have an influence on process automation. With a relevance value of 33 is the Worker overload the most important selection criterion for suitable RPA processes. The primary aim is to automate processes in which employee overloads and backlogs occur. In the opinion of interviewee (Intln) C, it is important to “create space for activities that add value, ie [...] focus on your own watch business” (C). The Availability of skilled workers often conflicts with the pressures and the scheduled capacities of an employee's line work. In order to be able to give feedback on RPA development on a regular basis, “above all, it must be clarified with [the] managers that these technical employees will also be released” (B).

When selecting a process, the added value for the customer should also be taken into account (Increase in customer satisfaction). This can be caused by direct effects, such as B. faster processing and shorter waiting times (A, C, D) or by indirect effects, such as the already explained relief of the employees (B, F) can be achieved. This gives clerks more time to “process the cases that really generate customer satisfaction” (D).

The Proposal of a department Another clear selection criterion for process optimization with RPA was mentioned by six interviewees. It is "extremely important [...] to include this personal feeling of employees in the process selection" (F) and to address the question through "direct discussions with the operational units" (E): "Which process disturbs the department and the Employees most? ”(F). The Acceptance of professionals regarded as a relevant criterion. Intln. F is of the opinion that "[...] those who have to work with the robot should [be considered] as the main factor, because robot processing always stands or falls with its acceptance". On the one hand, this could be due to the fact that employees develop fears about their job (A, F, J) and, on the other hand, “the employee's work process is changing” (A).

Qualitative process properties

Qualitative process properties group the criteria that embody properties of a suitable process, the process environment or process management, but which cannot be measured quantitatively in the sense of process indicators. With a relevance value of 26, the most important criterion in this dimension is die rule-based process control. An RPA use case should have a clear and rule-based process control (A, B, C, D, G, H, I, J, K) that can be transferred to an RPA bot. This then receives the ability to make decisions that “can be formulated sharply and in that sense also verifiable” (J). In addition, the processes should unite low level of complexity have. "The simpler the process, the easier it is for employees to be able to do without it" (F). Two experts also express their concerns, because "simple activities [are] also regarded as so-called relief activities" (H) for the employees. Exclusively complex activities could at some point lead to excessive demands (E).

Need RPA systems Access to structured data (B, C, H, G). Intl. I asks something like: “How do I get the necessary data? Are they available to me in a structured manner? ". Without access to structured data, automation is not possible or only possible with great effort (A, B, G, J). “I can have a PDF as an image or as a structured XML structure. I can read one of them very technically with a robot, for the other I need OCR, Optical Character Recognition, ”explains F. With a relevance value of 14, this is previous process optimization another relevant selection criterion. For this purpose, the statements refer either to optimizations that have already taken place (B) or to optimization potentials identified during the RPA development (A). “At best, the process has already been optimized […] and somehow […] standardized […]. Then it really makes sense to automate it. That is very important ", emphasizes B.

For editing the process should no expert intervention are required (A, B, D, I). In some internal practical cases, the only manual step is forwarding to the RPA bot by the employee, “because he has a better view of what the bot can and what [...] cannot. And then the bot actually does everything from start to finish, ideally […], so that the employee never sees him again ”(A).

Quantitative process properties

The quantitative process properties include properties of a process that can be clearly measured in the form of process indicators. With a relevance value of 27 is the Cost-benefit comparison the most important selection criterion of this dimension. An RPA use case should be profitable and increase efficiency (A, C, D, G, H, J). The experts often define the benefit on the basis of the achievable time savings of the acting bot compared to manual processing (A, C, D, G, J). One expert emphasizes that the possible use cases must be “viewed in the overall context” and, due to limited capacities, the effort and benefit between the use cases must be weighed (C). The evaluation of costs and benefits must also be followed up during automation (H). An expert reports on this: "In the past, everything that worked somehow was automated and [...] then [...] it was now so expensive to automate this business transaction that we could easily have just hired two employees" (D ). Almost all experts also mention the criterion Number of process instances. The process should "be carried out in a sufficiently large number [...]" (D) "[...] in order to be eligible for automation". A simple "process, which also occurs very often, [can] very quickly create a noticeable relief for the department" (F).C believes that "30, 40 cases a day [...], maybe 100 a week" should occur to get any benefit from RPA.

The Compensation of employee capacitiesAccording to the experts, which is achieved with the help of bots plays a central role, especially in the insurance concerned, as numerous employees have left the company in the past (C, D, G, H). Intln D formulates this as follows: “So the question no longer arises for us: Are we doing this in order to somehow save costs? That’s why we’re not doing it. Or do we just do it to add the volume of work? ”(D). Accordingly, the prioritization also focuses on use cases with a high MAK compensation (H).

The necessary Reduction of lead times is also a relevant criterion. According to Intln. C, RPA has the potential to "shorten process lead times". A robot is able to operate the user interfaces faster than a human (A, B, C, F, K). Only a longer system-related loading time can slow down the bot (F). Regardless of this, a robot is able to “start again immediately, while an employee naturally has [a] completely different reaction time […]” (F).

Risk and Compliance

The selection criteria that lead to compliance with laws and guidelines are in the dimension Risk and Compliance grouped, with the Compliance conformity the most important criterion is. Accordingly, processes are sorted out that cannot be implemented for compliance reasons (A). For example, "[...] no decisions may be made automatically at the expense of the customer" (A). To prevent the potential for fraud, two experts report an insecure use case, which should send payment instructions from outside to a robot by e-mail (A, K): “If I extract e-mails from mailboxes [...] somewhere and derive payments from them, that's a lot dangerous. I could just [...] send myself an email and then I would get € 100,000 ”(K). Ensuring data protection is relevant for the selection of processes. A data protection template is used to assess whether there are “basic concepts that are questionable under data protection law” (F).

The Avoiding Human Error includes another relevant selection criterion, because RPA is able to increase the accuracy and thus also the quality of the process processing in certain areas (C, D, I). Interview participant F describes this: "The fact of automation is that a robot simply processes its set of rules and makes the whole thing error-free". According to this, "it is worthwhile [...] precisely where [...] many mistakes can happen" (D).

The Avoidance of head monopolies when using RPA (A, B, C, D, F, H) is also mentioned frequently. Intln. A reports that the departments are already automating processes with the help of macros. At the same time, some employees in the departments are “extremely motivated” (F) and learn script languages ​​in order to use them in a professional context. “The problem is that there are always head monopolies” (F). As soon as the employee is no longer available, the solution “will soon no longer work because new software versions are being rolled out. Then nobody can adapt this tool ”(F). In order to displace this "shadow IT" (A, B), RPA represents a "sustainable solution" (A).

Information systems

The information systems dimension includes all selection criteria that relate to the existing process-relevant systems, their manual operation or the interfaces and their effects on automation. A relevant criterion addressed by all experts is the choice of one sensible bridge solution, since a clean alternative solution is not possible or too complex or needs to be compensated temporarily. Because “not everything can always be managed by a real IT project [...]. Most of the time, the resources are not available ”(A). Intln. K says: “Of course it would be more clever to marry off the existing systems […]. But since this host replacement is currently planned, we don't need to program anything in the host ”.

Closely related to a meaningful bridge solution is the ability to System interaction of the existing surfaces. "Especially in a heterogeneous architecture landscape, as is common in the insurance industry, [exist] systems that are cheaper [...] and [...] not so cheap compared to robot automation" (F). This is shown, for example, by the comparison between “software frameworks from the 1970s” (F) and a “web application that can be easily automated using a modern browser” (D).

Another relevant and at the same time controversial criterion is that Number of systemsthat a process suitable for RPA should include. On the one hand, the majority of experts recommend that a process that "ideally is marked by media discontinuities, ie where [the clerk] has to operate various IT systems in some form" (D) is suitable for automation with RPA. The benefit of the measure “potentially increases, the more systems are involved” (D), because existing automation using macros does not offer the possibility of processing cross-system processes (D, H). On the other hand, the fact that the complexity of the "entire development process, from business analysis to [...] maintenance and servicing" increases (C) speaks against a high number of systems involved.

Areas of responsibility and development

The dimension Areas of responsibility includes criteria that identify specific areas of application, tasks and areas that are particularly well suited for RPA processing. With a relevance score of 15 are above all repetitive tasks relevant for RPA. The automation of these tasks creates a relieving effect for the employees, since tasks “which have a certain monotonous character” (K) can be perceived by the employees as “tiring and almost degrading” (J). One of the experts provides an example for this in which "a steel price [has to be determined every day at 12 noon, which is then transferred to SAP" (K). Furthermore, should above all operational processes can be selected for automation (C, D, F, G, H). "I think that our operational areas [...] definitely represent the most potential, which is primarily due to the high [...] number of transactions and a high degree of standardization that already prevails" (F).

The relevant selection criterion Elaborate preparatory work comprises work steps that must be completed by an employee before the actual work can be started. "The clean routing of business transactions [...] to the right employees or to the right teams" (B) is an example of time-consuming work before the actual processing can begin. It is also described that bots can be used successfully to obtain and summarize relevant data from a wide variety of sources, so that the bot “not only supports the recipient with information and saves him the research effort, but [...] really takes away sub-processes from him” (I).

The potential, fast automations The only relevant criterion of the dimension is to be able to carry out and, accordingly, to prefer urgent use cases development. In this regard, an expert reports that the (current) management's claim is that corresponding measures must achieve results in the short term, "preferably this quarter" (A). Compared to more classic methods, which "take a lot of time until [...] the necessary developers are provided at all", the "development cycles [...] are extremely short" (J).

Discussion and outlook

If one compares the results of the interviews with the known criteria in the literature, it is noticeable that the criteria Rule-based process control, low degree of complexity, repetitive tasks, many process instances, compensation of employee capacities, digitally available structured data, optimized process and sensible bridge solution, are important for process selection both in the literature and in the interviews.

The criterion Transparent actual process costs is often given in the literature (Beetz and Riedl 2019; Smeets et al. 2019; Sutherland 2013; Syed et al. 2020), but not explicitly mentioned by the experts. Furthermore, Syed et al. (2020) the necessary Customer acceptance as well as the high data quality listed as additional criteria for the RPA input data, which is also not addressed by the experts questioned. Beetz and Riedl (2019) also name the Fault tolerance of a process that is to be processed automatically as a selection criterion.

The comparison of the criteria results in eight relevant criteria that have so far not been or insufficiently discussed in the literature. These include the criteria Employee overload, compliance conformity, ensured data protection, increase in customer satisfaction, system interaction, operational processes, avoidance of head monopolies and Elaborate preparatory work. They form the basis for further investigations.

Despite the methodologically sound approach (Vogelsang et al. 2013), the informative value of the results is limited. Only one group of companies and one consulting company were taken into account in the interviews. The strong expression of male interviewees can also influence the results. As in any qualitative-empirical work, the coding results are partly dependent on the coder's subjective perception. Building work should therefore validate the results achieved through further qualitative studies or through a survey in other organizations and industries or identify further criteria. In addition, the differences in the relevance of the individual criteria in different economic sectors and their influence on the success of the project are worth further consideration.


  1. van der Aalst WMP, Bichler M, Heinzl A (2018) Robotic process automation. Bus Inf Syst Eng 60: 269-272. https://doi.org/10.1007/s12599-018-0542-4

    Article Google Scholar

  2. Aguirre S, Rodriguez A (2017) Automation of a business process using robotic process automation (RPA): a case study. In: Figueroa-García et al (Ed) 4th Workshop on Engineering Applications, WEA. Springer, Berlin Heidelberg, pp 65-71

    Google Scholar

  3. Beetz R, Riedl Y (2019) Robotic process automation: developing a multi-criteria evaluation model for the selection of automatable business processes. 25th Americas Conference on Information Systems (AMCIS).

    Google Scholar

  4. Fung HP (2014) Criteria, use cases and effects of information technology process automation (ITPA). Adv Robot Autom. Https://doi.org/10.4172/2168-9695.1000124

    Article Google Scholar

  5. Lacity M, Willcocks L (2016) Robotic process automation at telefonica O2. MIS Q Executive 15 (1): 21-35

    Google Scholar

  6. Mayring P (2010) Qualitative content analysis. In: Mey G, Mruck K (Hrsg) Handbook Qualitative Research in Psychology. VS, Wiesbaden, pp. 601-613

    Google Scholar

  7. Smeets M, Erhard R, Kaußler T (2019) Robotic Process Automation (RPA) in the financial sector; Technology - implementation - success factors for decision-makers and users. Springer, Berlin Heidelberg

    Book Google Scholar

  8. Sutherland C (2013) Framing a constitution for robotistan; racing with the machine of robotic automation. HfS Research Ltd, https://www.hfsresearch.com/research/framing-constitution-robotistan/. Accessed: October 12, 2020.

    Google Scholar

  9. Syed R, Suriadi S, Adams M, Bandara W, Leemans SJJ, Ouyang C, ter Hofstede AHM, van de Weerd I, Wynn MT, Reijers HA (2020) Robotic process automation: contemporary themes and challenges. Comput Ind 115: 103162. https://doi.org/10.1016/j.compind.2019.103162

    Article Google Scholar

  10. Vogelsang K, Steinhueser M, Hoppe U (2013) A qualitative approach to examine technology acceptance. 34th International Conference on Information Systems (ICIS 2013).

    Google Scholar

  11. Willcocks L, Lacity M, Craig A (2015) The IT function and robotic process automation. https://www.uio.no/studier/emner/matnat/ifi/ITLED4021/h18/dokumentasjon/the-it-function-and-robotic-process-automation-willcocks-2015.pdf. Accessed June 24, 2020

Download references


Open Access funding enabled and organized by Projekt DEAL.

Author information


  1. Aachen University of Applied Sciences, Eupener Str. 70, 52066, Aachen, Germany

    Mathias Eggert

  2. Gothaer Systems GmbH, Gothaer Allee 1, 50969, Cologne, Germany

    Tobias Moulen

Corresponding author

Correspondence to Mathias Eggert.

Rights and permissions

Open Access This article is published under the Creative Commons Attribution 4.0 International License, which permits use, reproduction, editing, distribution and reproduction in any medium and format, provided you properly credit the original author (s) and source, a link to Include a Creative Commons license and indicate whether changes have been made.

The images and other third-party material contained in this article are also subject to the named Creative Commons license, unless otherwise stated in the legend. If the material in question is not under the named Creative Commons license and the action in question is not permitted under statutory provisions, the consent of the respective rights holder must be obtained for the further uses of the material listed above.

For more details on the license, please refer to the license information at http://creativecommons.org/licenses/by/4.0/deed.de.

Reprints and Permissions

About this article

Cite this article

Eggert, M., Moulen, T. Selection of business processes for the application of robotic process automation using the example of an insurance company. HMD57, 1150-1162 (2020). https://doi.org/10.1365/s40702-020-00665-0

Download citation


  • Robotic Process Automation
  • RPA
  • Process automation
  • Process improvement
  • Selection criteria
  • Process selection


  • Robotic Process Automation
  • RPA
  • Process automation
  • Process improvement
  • Selection criteria
  • Process selection