REVIEW
Robotic Process Automation - A Systematic Mapping Study and
Classification Framework
Judith Wewerka 0000-0002-4809-2480 a,b
and
Manfred Reichert 0000-0003-2536-4153 a
aInstitute of Databases and Information Systems, Ulm University, Germany
bResearch and Development, BMW Group Munich, Germany
{judith.wewerka, manfred.reichert}@uni-ulm.de
ARTICLE HISTORY
Compiled October 8, 2021
ABSTRACT
Robotic Process Automation (RPA) deals with the automation of rule-based process
tasks to increase process efficiency and to reduce process costs. Due to the utmost
importance of business process automation in industry, RPA attracts increasing at-
tention in the scientific field as well. This paper presents the state-of-the-art in the
RPA field by means of a Systematic Mapping Study (SMS). In this SMS, 63 publica-
tions are identified, categorised, and analysed along well-defined research questions.
From the SMS findings, additionally, a framework for systematically analysing, as-
sessing, and comparing existing as well as upcoming RPA works is derived. The
discovered thematic clusters suggest further investigations in order to develop a
more elaborated structural research approach for RPA.
KEYWORDS
Systematic Mapping Study; Robotic Process Automation; RPA; Systematic
Evaluation; RPA Classification Framework
1. Introduction
In our continuously changing world, it is indispensable that business processes are
highly adaptive (Reichert and Weber 2012) and become more efficient and cost-
effective (Lohrmann and Reichert 2016). As a consequence, companies demand for
an increasing degree of business process automation to stay competitive in their mar-
kets. In this context, business process management (BPM) plays a crucial role in
the digital lifecycle support of business processes involving multiple participants and
software systems (Weber, Sadiq, and Reichert 2009). Currently, BPM is enhanced by
the use of process mining and Robotic Process Automation (RPA). While the for-
mer gives companies objective and data-driven insights into the actual flow of their
business processes (van der Aalst 2011), the latter describes software robots (bots
for short) mimicking human interaction (Asatiani and Penttinen 2016). RPA consti-
tutes a ‘highly promising approach’ (Cewe, Koch, and Mertens 2017) and more and
more companies rely on this cutting edge technology (Asatiani and Penttinen 2016)
to optimise, implement, and automate selected process tasks.
1.1. Problem Statement
RPA constitutes an emerging technology raising high expectations in industry (Auth
and Bensberg 2019). For enterprises, however, it is still difficult to grasp the funda-
mental concepts of RPA, to understand the differences in comparison to other methods
and technologies (e.g., BPM and process mining), and to estimate the effects the in-
troduction of RPA will have on an enterprise and its staff. As preliminary work and
preparation of this systematic mapping study (SMS), we conducted an exploratory
case study in the automotive industry (Wewerka and Reichert 2021), whose results
motivate the problem described above.
Due to the increasing scientific attention of RPA (see, for example, the RPA Fo-
rum established in conjunction with the BPM Conference Series1), the number of
publications on RPA will further increase over time. RPA projects have an impact
on organisation and management, IT enterprise architectures, business processes, and
business process stakeholders (Auth and Bensberg 2019; Fernandez and Aman 2018;
Ian, Dhayalan, and Andy 2016). Therefore, RPA publications can be attributed to
different scientific areas, including e.g., management or computer science. As opposed
to other SLRs in the field (e.g., (Ivanˇci´c, Vugec, and Vukˇsi´c 2019; Riedl and Beetz
2019)), this work considers publications from these different fields in order to provide
a holistic overview of RPA publications, not just focusing on single aspects such as or-
ganisational or socio-technical ones. More precisely, we conducted an SMS to provide
a structured overview of RPA research. The results of this study, in turn, have been
used to systematically derive a classification framework for analysing, assessing, and
comparing RPA works (Petersen, Vakkalanka, and Kuzniarz 2015). In particular, the
framework shall help researchers to easily find relevant RPA publications.
1.2. Contribution
This paper provides an SMS on RPA as well as a classification framework derived
from it. We aim to present the state-of-the-art on RPA by systematically analysing
and assessing the most relevant publications in the field. In this context, we provide
meanings attached to RPA, discuss differences to related technologies, introduce crite-
ria for RPA-suitable process tasks, and give insights into RPA effects. Furthermore, we
present case studies, give an overview of RPA methods, and discuss the combination
of Artificial Intelligence (AI) methods with RPA. Taking the results of the SMS, we
derive the ANCOPUR classification framework for systematically mapping, analysing,
and comparing emerging publications on RPA, and link them to existing publications
to assess their novelty and research contribution. The main contributions of our work
to RPA research are as follows:
1. We provide a holistic overview of the state-of-the-art of RPA research that cov-
ers multiple perspectives, including organisation and management, IT enterprise
architectures, business processes, and business process stakeholders.
2. We focus on seven thematic RPA clusters: RPA meanings, differences to related
technologies, criteria for selecting process tasks, RPA use cases, RPA effects,
methods for RPA projects, and RPA and AI.
3. We derive a classification framework to analyse, assess, and compare RPA re-
search works.
4. We categorise methods for RPA projects along the software development life
1https://bpm2021.diag.uniroma1.it/call-for-rpa-forum/
2
cycle stages.
The remainder of this paper is structured as follows. First, backgrounds on RPA
and related work are provided in Section 2. Section 3 introduces the research method-
ology we applied, followed by the obtained results in Section 4. Section 5 derives the
ANCOPUR classification framework. Then, the results of this article are discussed in
Section 6. We conclude with a summary and an outlook in Section 7.
2. Backgrounds
This section presents background information needed for understanding this work.
First, Robotic Process Automation is explained and an example is given. Second,
related work from the RPA field is summarised. To start, the differences between
business process management, process mining, and RPA are clarified (cf. Figure 1).
BPM deals with the modelling, implementation, execution, monitoring, and evolution
of business processes (Reichert and Weber 2012). Process mining, in turn, may be
considered as a sub-discipline of BPM that supports business process discovery (i.e.,
to discover business process models from event logs), conformance checking (i.e., to
check to what degree a given event log and business process model conform with
each other), and data-driven business process analysis (e.g., log-based verification of
business process compliance) (Geyer-Klingeberg, Nakladal, and Baldauf 2018). Finally,
RPA targets at the automation of process tasks and, thus, hands them over to a bot.
All three disciplines can be induced by artificial intelligence.
Process mining can also help to discover process models from the interactions a
process participant performs with software systems and, thus, seems to be appropriate
for identifying process tasks being suitable for RPA (1). In addition, process mining can
analyse RPA logs and, thus, monitor RPA bots (1). RPA supports BPM by enabling
full automation of selected process tasks and contributes to improved business process
performance (3). Process mining, in particular business process discovery techniques
provide valuable input for BPM projects, e.g., the discovery of the actual as-is business
process (2). Recently, process discovery has increasingly been used to derive RPA
scripts for routine process tasks from user interface (UI) logs (1).
Robotic
Process
Automation
Process
Mining
Business
Process
Management
1
23
Figure 1. Relation between Process Mining, Business Process Management, and Robotic Process Automa-
tion.
2.1. Robotic Process Automation
The IEEE Standards Association defines RPA as follows (IEEE 2017): ‘A preconfig-
ured software instance that uses business rules and predefined activity choreography
to complete the autonomous execution of a combination of processes, activities, trans-
3
actions, and tasks in one or more unrelated software systems to deliver a result or
service with human exception management.’
For a better understanding we consider an example (Aguirre and Rodriguez 2017).
Note that we distinguish between the notions of business process and process task. The
former consists of a set of activities (i.e., process tasks) that are performed in coordina-
tion in an organisational and technical environment. These process tasks jointly realise
a business goal (Weske 2007), and describe pieces of work to be performed within a
certain time period (Mundbrod and Reichert 2017). RPA, in turn, automates one or
several tasks of a business process. We illustrate this relation by an example: Figure 2
visualises the business process of a payment generation: the customer communicates
with the call centre and requests a payment receipt. In the front office, the agent re-
ceives the request and creates the corresponding case in the Customer Relationship
Management (CRM) database. A back office agent opens the case from the database,
generates the payment receipt from the accounting system, and sends the receipt via
email to the customer. Finally, the case is closed in the CRM database. These process
tasks of the back office agent are rule-based, highly repetitive, and well-structured.
Furthermore, two unrelated software systems, i.e., the CRM and the accounting sys-
tems are used. A correlation between these two systems can be established via the
customer ID, which is copied and pasted from one system to the other. According to
the aforementioned IEEE definition, RPA could be applied to deliver the same result.
Generate payment receipt Customer
Front office
agent
Back office
agent
Manual
process
tasks
Communicate
with the call
center
Request
payment receipt
Receive request
and create the
case
Open the case Generate the
payment receipt
Send the
payment receipt
(e-mail)
Close the case
Receive the
payment receipt
(e-mail)
CRM Accounting
system CRM
Payment receipt request
Payment receipt
Figure 2. Business Process of Payment Generation without RPA based on (Aguirre and Rodriguez 2017)
.
After applying RPA, the business process remains the same. The four process tasks
that have been manually performed by the back office agent so far, are now replaced
by one automated task implemented with RPA (cf. Figure 3). The bot accesses the
CRM system, copies the customer ID, and pastes it to the accounting system. Then,
the corresponding payment receipt is generated, and the bot sends it to the customer
before closing the case.
2.2. Related Work
Systematic literature reviews (SLRs) and Systematic Mapping Studies (SMSs) ex-
ist in various areas. Especially, they are used to cluster research on emerging tech-
nologies, e.g., big data analytics (Khanra et al. 2020), Internet of Things (Ng et al.
4
Generate payment receipt Customer
Front office
agent
RPA bot
Auto-
mated
process
task
Communicate
with the call
center
Request
payment receipt
Receive request
and create the
case
Take
information
from CRM
Generate the
payment receipt
Send the
payment receipt
(e-mail)
Close the case
Receive the
payment receipt
(e-mail)
CRM Accounting
system CRM
Payment receipt request
Payment receipt
Figure 3. Business Process of Payment Generation with RPA for some process tasks based on (Aguirre and
Rodriguez 2017).
2018), business process management software (Steinau et al. 2019), or process mining
(El-Khawaga et al. 2020). This section gives a short overview of literature research
approaches in the RPA area.
To the best of our knowledge, there is only one other SMS (Enriquez et al. 2020)
analysing the current state-of-the-art of RPA. The main focus of this SMS, is to
evaluate 14 commercial RPA tools concerning the coverage of 48 functions mapped to
RPA life cycle phases. As major result, the operation phase is covered by over 80%
of the RPA tools, whereas support for the analysis phase is below 15%. This shows
the potential for process mining when applying it to UI event logs. Note that this
SMS focuses on the technology perspective of RPA, while our SMS takes a holistic
approach.
Furthermore, there exist SLRs dealing with RPA. Note that SMS and SLR differ
in several aspects, e.g., an SMS addresses multiple and broader research questions
and data analysis is descriptive summarising existing data, whereas an SLR uses more
in-depth analysis techniques. Moreover, the main goal of an SMS is to provide an
overview of the scope of the research area (Kitchenham and Charters 2007; Petersen,
Vakkalanka, and Kuzniarz 2015). We are aware of the following SLRs on RPA:
•(Riedl and Beetz 2019): The main focus of this SLR is to derive selection cri-
teria for assessing the RPA suitability of process tasks as well as to develop a
corresponding evaluation method.
•(Ivanˇci´c, Vugec, and Vukˇsi´c 2019): This SLR aims to systematically investigate
RPA experiences from business practices in Scopus and Web of Science.
•(Syed et al. 2020): RPA-related topics and challenges for future research are
investigated. This SLR focuses on the description of RPA readiness and maturity,
the potential of RPA, an effective RPA methodology, and RPA technologies. The
results of the SLR are used to highlight key research challenges for future RPA
research.
Finally, we discovered two works relying on a literature review and other methods
to address research questions in the RPA area:
•(Gotthardt et al. 2019) examines the current state of RPA as well as fundamental
challenges in accounting and auditing. For this purpose, a literature review,
5
interview results, and case studies are presented to summarise key factors. In
particular, (Gotthardt et al. 2019) follows a domain-specific approach by focusing
on accounting and auditing, with a special emphasis on the role of AI.
•(Santos, Pereira, and Vasconcelos 2019) provides an approach for evaluating RPA
development in enterprises. A conceptual model on the relationships between
RPA topics, identified in a literature review, is presented. The model consists of
three steps, i.e., definition of strategic goals, process task assessment, and tactical
evaluation as well as factors for a successful RPA implementation. Influencing
factors include benefits, disadvantages, selection criteria, future challenges, and
future opportunities.
In summary, to the best of our knowledge, there are no other publication addressing
the scope presented in Section 1.1.
3. Methodology
We conduct an SMS to provide an overview of RPA research from the perspectives
of organisation and management, IT enterprise architectures, business processes, and
business process stakeholders as well as to structure this research area (Kitchenham
and Charters 2007; Petersen, Vakkalanka, and Kuzniarz 2015). Following the guidelines
described by Petersen as well as the procedures suggested by (Khanra et al. 2020; Ng
et al. 2018), we design a protocol (cf. Figure 4) that describes the formulation of
research questions (cf. Section 3.1), the definition of rules for conducting the search
(cf. Section 3.2), the selection of publications (cf. Section 3.3), the data extraction
method (cf. Section 3.4), and the data analysis method (cf. Section 3.5).
3.1. Formulation of the Research Questions
Our general goal is to analyse the body of relevant publications in the RPA field. There-
fore, we opt for multidisciplinarity and investigate RPA from different perspectives. In
a first step, we want to understand the technology perspective of RPA and how it dif-
fers from related technologies, like intelligent automation (Bruno, Johnson, and Hesley
2017; Schmitz, Stummer, and Michael 2019; Suri et al. 2018) or BPM (Cewe, Koch,
and Mertens 2017). Additionally, the exploratory case study performed in preparation
to our SMS identified the challenge to explain RPA to end users (Wewerka and Re-
ichert 2021). This results in our first research question: RQ 1: What meanings are
attached to RPA in literature and what are the differences between RPA
and related technologies? Secondly, managers want to understand what can be
automated. Hence, criteria for assessing whether or not given process tasks are suited
for RPA are investigated. One major challenge in RPA projects is therefore to iden-
tify the business process tasks suited for automation (Wewerka and Reichert 2021).
Furthermore, from a technology perspective we are interested in the tools available
for implementing RPA. This leads to our second research question: RQ 2: Which
process tasks can be automated with RPA and which tools are used for au-
tomation? For newly emerging technologies, like RPA, the question arises whether
it is worthwhile to adopt it. Therefore, from both an organisational and an economic
perspective, we want to systematically understand RPA effects on humans and their
daily worklife as well as on the enterprises implementing RPA projects. This results in
our third research question: RQ 3: What are RPA effects? In a fourth step, we in-
6
Specific research question
of intellectual interest
1. Formulation of Research
Questions
(cf. Section 3.1)
3. Selection of Publications
(cf. Section 3.3)
Search String: `robotic process automation' OR `intelligent process
automation' OR `tools process automation' OR `artificial intelligence in
business process' OR `machine learning in business process' OR `cognitive
process automation'
Data Sources: ACM Digital Library, Science Direct – Elsevier, IEEE Xplore
Digital Library, SpringerLink, Google Scholar
Inclusion criteria: 1510 publications identifed
Exclusion criteria:
- Eliminate duplicates and publications not written in English (465 excluded)
- Eliminate publications whose title do not fit our research (756 excluded)
- Eliminate publications whose abstract do not fit our research (88 excluded)
- Eliminate patents, theses, or web pages (59 excluded)
- Elminate non accessible publications (17 excluded)
- Eliminate publications included in previous research (40 excluded)
- Elminate publications without new input (46 excluded)
Backward referencing: 1 publication added
Alerts: 23 publications added
1. General information
2. Meanings attached to RPA
3. Differences between RPA and related technologies
4. Criteria for selecting suitable process tasks for RPA
5. Concrete process tasks automated in specific business areas with an
explicitly mentioned automation tool
6. RPA effects on humans, worklife, and companies
7. Methods to improve RPA projects
8. Combination of RPA with AI
9. Significant information outside the scope of the derived RQ
Comprehensive
understanding and future
research agendas
RQ 1: What meanings are attached to RPA in literature and what are the
differences between RPA and related technologies?
RQ 2: Which process tasks can be automated with RPA and which tools are
used for automation?
RQ 3: What are RPA effects?
RQ 4: Are there methods for improving the implementation of RPA projects?
RQ 5: Is AI used in combination with RPA?
2. Definition of Search Rules
(cf. Section 3.2)
4. Extraction of Data
(cf. Section 3.4)
5. Analysis of Data
(cf. Section 3.5)
Cluster data based on matches and differences
Figure 4. Protocol for Systematic Mapping Study (inspired by (Khanra et al. 2020)).
vestigate how far research has taken up on RPA. Note that all aforementioned aspects
only work with methodological support. Particularly, we are interested in methods
that aim to foster RPA implementation. The exploratory case study revealed that one
particular challenge is to provide software development guidelines for RPA projects
(Wewerka and Reichert 2021). This leads to our fourth research question: RQ 4: Are
there methods for improving the implementation of RPA projects? Finally,
the growing importance of AI in many areas of automation raises the question to what
degree AI plays a role in connection with business process automation. The fifth re-
search question addresses the topic of combining AI with RPA: RQ 5: Is AI used
in combination with RPA?
3.2. Definition of Search Rules
Search String. We elaborate the search string iteratively based on our expert knowl-
edge of the topic, the pre-specified research questions, and pilot searches. The search
string is refined to retrieve a maximum number of publications to meet the goal of the
SMS, i.e., to broadly cover the research area RPA from the aforementioned perspec-
tives (organisation and management, IT enterprise architectures, business processes,
and business process stakeholders) (Kitchenham and Charters 2007). The final search
string for the SMS is as follows:
‘robotic process automation’ OR ‘intelligent process automation’ OR
7
‘tools process automation’ OR ‘artificial intelligence in business process’
OR ‘machine learning in business process’ OR ‘cognitive process
automation’.
Note that the acronym ‘RPA’ is not included, as the search then would yield around
31.000 results. RPA not only serves as acronym for Robotic Process Automation, but
also for Recombinase Polymerase Amplification in the field of DNA chemistry and
others. Though we omit RPA in the search string, all relevant publications are still
included in the results.
Data Sources. We apply the search string to different data sources to discover
relevant publications. Five electronic libraries are identified as relevant for conducting
the SMS as they cover scientific publications in Computer Science, Management, and
IT enterprise architecture:
ACM Digital Library, Science Direct - Elsevier, IEEE
Xplore Digital Library, SpringerLink, and Google Scholar.
Additionally, we consider literature cited by the retrieved publications by performing
abackward reference search (Jalali and Wohlin 2012). Finally, Google Scholar alerts
are analysed during both the SMS procedure and the writing process to get notified
about newly emerging publications on the topic.
Inclusion and Exclusion Criteria. To identify relevant publications, we define
the following inclusion and exclusion criteria.
Inclusion Criteria:
1.) The publication deals with the topic of RPA and contributes answers to at least
one of the aforementioned research questions.
2.) The title and the abstract seem to contribute to our research questions and
contain terms such as robotic/intelligent/cognitive process automation, virtual
assistant, process intelligence, business process model automation, intelligent
business process management, or software bot.
Exclusion Criteria:
1.) The publication is not written in English.
2.) The title and abstract do not seem to contribute to our research questions and
contain terms such as business process management, business intelligence, ana-
lytics, multi-agent system, big data, or process mining.
3.) The publication is a patent, master thesis, or web page.
4.) The publication is not electronically accessible without payment.
5.) All relevant aspects of the publication are included in another publication.
6.) The publication only compares existing research and has no new input.
A publication is included if the inclusion criteria are met, but is then excluded if
any of the exclusion criteria is fulfilled.
3.3. Selection of Publications
The search string is applied to the identified data sources, which yields 1510 results
(Inclusion Criterion 1). To select relevant publications, the metadata is loaded into
Microsoft Excel. It includes title, author, year, abstract, and keywords. In a first step,
duplicates and publications not written in English (Exclusion Criterion 1) are excluded
resulting in 1045 publications afterwards. Then, publications whose title does not in-
8
dicate any contribution to one of the research questions are excluded, leaving 289 pub-
lications (Inclusion Criterion 2, Exclusion Criterion 2). Following this, the abstracts of
the remaining publications are scanned leading to 201 publications (Inclusion Crite-
rion 2, Exclusion Criterion 2). We then exclude publications corresponding to patents,
theses, or web pages, resulting in 142 relevant publications (Exclusion Criterion 3).
Thereof, 125 are accessible without payment (Exclusion Criterion 4) and 85 are not
included in another publication (Exclusion Criterion 5). Finally, 39 publications pro-
vide new input to the research questions and are included in the final publication list
(Exclusion Criterion 6). Through backward referencing one additional publication is
identified and included.
The initial search was performed on 6 June 2019. Since then (until June 2020) the
alerts from Google Scholar have revealed 1206 new publications. 23 of them meeting
the inclusion criteria and not fulfilling any of the exclusion criteria. Thus, they are
added to our final publication list, which comprises 63 relevant publications in total.
3.4. Data Extraction Method
To each of the 63 relevant publications, a data extraction method is applied in order
to answer the research questions derived in Section 3.1. We extract the following
information:
1.) General information, i.e., title, author, publication year, publication venue, num-
ber of citations, and publication type,
2.) Meanings attached to RPA (RQ 1),
3.) Differences between RPA and related technologies, e.g., intelligent automation,
BPM, etc. (RQ 1),
4.) Criteria for selecting suitable process tasks for RPA (RQ 2),
5.) Concrete process tasks automated in specific business areas with an explicitly
mentioned automation tool (RQ 2),
6.) RPA effects on humans, worklife, and enterprises (RQ 3),
7.) Methods to improve RPA projects (RQ 4),
8.) Combination of RPA with AI (RQ 5), and
9.) Significant information outside the scope of the derived research questions.
Tables 1 and 2 give an overview of the 63 relevant publications indicating the refer-
ence, ID, title, type of publication, and research questions the publication refers to. In
the following, the ID is used to refer to the corresponding publication. The publication
type, in turn, distinguishes between Method,Case Study,Review, and Research Paper.
A publication is classified as Method if it reports on the development and testing of a
new RPA method, as Case Study if it focuses on a practical use case, as Review if it
provides a synthesis of acquainted knowledge, and as Research otherwise.
3.5. Data Analysis Method
After having extracted relevant data from all selected publications, we cluster the
obtained data. For each research question, we scan relevant information and build
groups based on matches and differences.
Concerning RQ 1, we study all meanings attached to RPA by the publications,
identify different aspects, e.g., ‘software-based solution’, ‘mimics human behaviour’ or
‘rule-based nature’, and label the publications according to the aspects they cover.
9
Table 1. Final list of the 63 relevant publications indicating reference, ID, and answers to the research questions.
Ref ID Title Type RQ1 RQ2 RQ3 RQ4 RQ5
(Aguirre and Rodriguez 2017) P01 Automation of a Business Process Using Robotic Process Automation (RPA):
A Case Study
Case Study x x x
(Asatiani and Penttinen 2016) P02 Turning robotic process automation into commercial success - Case Opus-
Capita
Research x x x
(Asquith and Horsman 2019) P03 Let the robots do it! - Taking a look at Robotic Process Automation and its
potential application in digital forensics
Case Study x x x
(Auth and Bensberg 2019) P04 Impact of Robotic Process Automation on Enterprise Architectures Research x
(Bosco et al. 2019) P05 Discovering Automatable Routines From User Interaction Logs Method x
(Bruno, Johnson, and Hesley 2017) P06 Robotic disruption and the new revenue cycle Research x x x
(Cewe, Koch, and Mertens 2017) P07 Minimal Effort Requirements Engineering for Robotic Process Automation
with Test Driven Development and Screen Recording
Method x x
(Chac´on-Montero, Jim´enez-Ram´ırez,
and Enr´ıquez 2019)
P08 Towards a Method for Automated Testing in Robotic Process Automation
Projects
Method x
(Chalmers 2018) P09 Machine Learning with Certainty: A Requirement For Intelligent Process Au-
tomation
Research x
(Cohen, Rozario, and Zhang 2019) P10 Exploring the Use of Robotic Process Automation (RPA) in Substantive Au-
dit Procedures
Case Study x x x
(Eikebrokk and Olsen 2019) P11 Robotic Process Automation for Knowledge Workers - Will It Lead To Em-
powerment or Lay-Offs?
Research x
(Eikebrokk and Olsen 2020) P12 Robotic Process Automation and Consequences for Knowledge Workers; a
Mixed-Method Study
Research x
(Fernandez and Aman 2018) P13 Impacts of Robotic Process Automation on Global Accounting Services Research x
(Fung 2014) P14 Criteria, Use Cases and Effects of Information Technology Process Automa-
tion (ITPA)
Research x x x
(Gao et al. 2019) P15 Automated robotic process automation: A self-learning approach Method x
(Geyer-Klingeberg, Nakladal, and Bal-
dauf 2018)
P16 Process Mining and Robotic Process Automation: A Perfect Match Method x
(Hallikainen, Bekkhus, and Pan 2018) P17 How OpusCapita Used Internal RPA Capabilities to Offer Services to Clients Case Study x x
(Hindel, Cabrera, and Stierle 2020) P18 Robotic Process Automation: Hype or Hope? Research x
(Houy, Hamberg, and Fettke 2019) P19 Robotic Process Automation in Public Administrations Research x x x
(Huang and Vasarhelyi 2019) P20 Applying robotic process automation (RPA) in auditing: A framework Method x
(Hwang et al. 2020) P21 MIORPA: Middleware System for open-source robotic process automation Method x
(Ian, Dhayalan, and Andy 2016) P22 The future of professional work: Will you be replaced or will you be sitting
next to a robot?
Method x
(Issac, Muni, and Desai 2018) P23 Delineated Analysis of Robotic Process Automation Tools Review x
(Jim´enez-Ram´ırez et al. 2020) P24 Automated testing in robotic process automation projects Method x
(Jim´enez-Ram´ırez et al. 2019) P25 A Method to Improve the Early Stages of the Robotic Process Automation
Lifecycle
Method x x
(Kin et al. 2018) P26 Cognitive Automation Robots (CAR) Research x
(Koch et al. 2020) P27 ‘Mirror, Mirror, on the wall’: Robotic Process Automation in the Public Sector
using a Digital Twin
Method x
(Kokina and Blanchette 2019) P28 Early evidence of digital labor in accounting: Innovation with Robotic Process
Automation
Research x x
(Lacity and Willcocks 2017) P29 A New Approach to Automating Services Research x x
(Lacity, Willcocks, and Craig 2015a) P30 Robotic process automation at Xchanging Case Study x x
(Lacity, Willcocks, and Craig 2015b) P31 Robotic Process Automation: Mature Capabilities in the Energy Sector Case Study x x
(Lacity, Willcocks, and Craig 2016) P32 Robotizing Global Financial Shared Services at Royal DSM Case Study x x
(Lacity, Willcocks, and Craig 2017) P33 Service Automation: Cognitive Virtual Agents at SEB Bank Research x x x
(Leno et al. 2020a) P34 Automated Discovery of Data Transformations for Robotic Process Automa-
tion
Method x
10
Table 2. Final list of the 63 relevant publications indicating reference, ID, and answers to the research questions - Continuation.
Ref ID Title Type RQ1 RQ2 RQ3 RQ4 RQ5
(Leno et al. 2018) P35 Multi-Perspective process model discovery for robotic process automation Method x x
(Leno et al. 2020b) P36 Robotic Process Mining: Vision and Challenges Method x
(Leno et al. 2019) P37 Action Logger: Enabling Process Mining for Robotic Process Automation Method x
(Leopold, van Der Aa, and Reijers
2018)
P38 Identifying candidate tasks for robotic process automation in textual process
descriptions
Method x
(Leshob, Bourgouin, and Renard 2018) P39 Towards a Process Analysis Approach to Adopt Robotic Process Automation Method x
(Lewicki, Tochowicz, and van
Genuchten 2019)
P40 Are Robots Taking Our Jobs? A RoboPlatform at a Bank Case Study x x x
(Masood and Hashmi 2019) P41 Cognitive Robotics Process Automation: Automate This! Research x x
(Moffitt, Rozario, and Vasarhelyi
2018)
P42 Robotic Process Automation for Auditing Research x
(Mohanty and Vyas 2018) P43 Intelligent Process Automation = RPA + AI Research x x x
(Osmundsen, Iden, and Bygstad 2019) P44 Organizing Robotic Process Automation: Balancing Loose and Tight Cou-
pling
Method x x
(Patel et al. 2019) P45 Customized Automated Email Response Bot using Machine Learning and
Robotic Process Automation
Research x x
(Penttinen, Kasslin, and Asatiani
2018)
P46 How to Choose Between Robotic Process Automation and Back-End System
Automation?
Research x x x
(Radke, Dang, and Tan 2020) P47 Using Robotic Process Automation (RPA) to enhance Item Master Data
Maintenance Process
Case Study x x
(Riedl and Beetz 2019) P48 Robotic Process Automation: Developing a Multi-Criteria Evaluation Model
for the Selection of Automatable Business Processes
Method x
(Rutschi and Dibbern 2020) P49 Towards a framework of implementing software robots: Transforming Human-
executed Routines into Machines
Method x
(Schmitz, Dietze, and Czarnecki 2019) P50 Enabling digital transformation through robotic process automation at
Deutsche Telekom
Case Study x x
(Schmitz, Stummer, and Michael 2019) P51 Smart Automation as Enabler of Digitalization? A Review of RPA/AI Poten-
tial and Barriers to Its Realization
Review x x x
(S´eguin and Benkala¨ı 2020) P52 Robotic Process Automation (RPA) Using an Integer Linear Programming
Formulation
Method x
(Stople et al. 2017) P53 Lightweight IT and the IT Function: experiences from robotic process au-
tomation in a Norwegian bank
Case Study x x
(Suri et al. 2018) P54 Automation of Knowledge-Based Shared Services and Centers of Expertise Research x x
(Suri, Elia, and van Hillegersberg
2017)
P55 Software Bots - The next frontier for shared services and functional excellence Research x
(van der Aalst, Bichler, and Heinzl
2018)
P56 Robotic Process Automation Research x x
(Wanner et al. 2020) P57 Process selection in RPA projects - Towards a quantifiable method of decision
making
Method x
(Willcocks and Lacity 2015) P58 Robotic Process Automation: The Next Transformation Lever for Shared Ser-
vices
Case Study x x x
(Willcocks and Lacity 2016) P59 Robotic Process Automation at Telef´onica O2 Case Study x x x
(Willcocks, Lacity, and Craig 2015) P60 The IT Function and Robotic Process Automation Research x
(William and William 2019) P61 Improving Corporate Secretary Productivity Using Robotic Process Automa-
tion
Case Study x x
(Wr´oblewska et al. 2018) P62 Robotic Process Automation of Unstructured Data with Machine Learning Research x x x
(Yatskiv et al. 2019) P63 Improved Method of Software Automation Testing Based on the Robotic
Process Automation Technology
Case Study x
11
The same procedure is applied to bundle differences to other technologies (RQ 1),
criteria for selecting process tasks (RQ 2), and RPA effects (RQ 3).
Depending on the publication type, different data analysis methods are applied.
Concerning case studies, we investigate the business area, the concerned process task,
and the used automation tool. Then, we cluster these case studies (RQ 2). Method
papers are co-related with the stage of the RPA project they aim to improve, in order
to identify common points (RQ 4). Finally, research papers answering RQ 5 are treated
separately to cluster approaches that combine RPA with AI.
The data analysis method aims to facilitate the derivation of a classification frame-
work from our SMS results (Petersen et al. 2008).
4. Results
We analyse the 63 publications discovered with the SMS to answer the research ques-
tions (cf. Section 3.1). The answers are structured along the research questions and
the seven discovered thematic clusters (i.e., RPA meanings, differences to related tech-
nologies, criteria for selecting process tasks, RPA use cases, RPA effects, methods for
RPA projects, and RPA and AI). Table 3 gives an overview on which section covers
which research question.
Table 3. Overview of Result Section.
Section Content Page
4.1 RPA meanings 12
Differences to related technologies 13
4.2 Criteria for selecting process tasks 15
RPA use cases 16
4.3 RPA effects 17
4.4 Methods for RPA projects 19
4.5 RPA and AI 23
We have noticed a growing interest in RPA in the scientific literature. Figure 5
shows the distribution of the publications included in this SMS over the recent years;
it started with one to seven publications in the years 2014 to 2017. In 2018, 15 relevant
publications appeared and in 2019, 21 works were published. In 2020 (until June), 11
publications could be identified.
Concerning the publication venue, there is no clear majority. RPA is important in a
variety of areas covered by different conferences and journals. Regarding authorship,
two researchers are dominating: M. Lacity and L. Willcocks are both co-authors of
eight publications each.
4.1. RQ 1: What meanings are attached to RPA in literature and what
are the differences between RPA and related technologies?
In Section 2 we provided the IEEE definition for RPA. The meanings attached to
RPA in literature are revealed in this section. Further, we discuss differences to re-
lated technologies. It is important to profoundly understand RPA in order to avoid
confusion with similar technologies. A clear positioning of the RPA technology between
robotic desktop automation, intelligent automation, and BPM is, therefore, desirable.
Remember that RPA aims to automate selected process tasks of business processes.
RPA Meanings. A first definition of RPA in literature can be found in P60: ‘RPA
12
0
5
10
15
20
2014 2015 2016 2017 2018 2019 2020 (until
June)
Number of Publications
Year
Figure 3: Distribution of Publications over Years.
Figure 5. Distribution of Publications over Years.
is a software-based solution [...] [and] refers to configuring the software “robot” to
do the work previously done by people.’ This definition addresses two aspects. First,
RPA corresponds to a software-based solution (cf. P33, P51, P58, P59). Second,
it mimics human behaviour (cf. P43, P45, P51, P53, P59). Most publications pick
up these meanings expanding them by two other characteristics (i.e., task automation
and non-invasiveness). Instead of ‘software-based solution’, terms like ‘software robot’
(P40, P43) or ‘virtual assistant’ (P02) are used as well. Mimicking human behaviour
is also expressed by phrases like ‘enters data, just as a human would’ (P40), ‘mimics
human actions’ (P06), or ‘operates [...] in the way a human would do’ (P56).
Characteristics of the routines (i.e., process tasks) automated by software robots
are included in the definitions. These characteristics cover the rule-based nature of
the routines (P01, P10, P29, P45), the inclusion of structured data (P01, P10, P29,
P53), and the emphasis on routine tasks (P01, P03, P06, P10, P51). Furthermore,
some publications emphasise the non-invasiveness of RPA, meaning that RPA does
not change the involved software systems (P06, P46, P51), but automation is realised
on top of them. Figure 6 summarises the RPA meanings we discovered in the SMS.
Note that these meanings also help to position RPA in relation to other approaches
and technologies (see below).
RPA Meanings
Software-based solution
Mimic human behaviour
Uses structured data
Non-invasive
Routine tasks
Automate rule-based process tasks
Figure 6. Meanings attached to RPA in literature.
Differences between RPA and Related Technologies. In the following, we
analyse the differences between RPA and Robotic Desktop Automation (RDA),Intel-
ligent/Cognitive RPA, and Business Process Management (BPM). These terms are
mentioned the most frequently in the results of the SMS.
As major difference between RDA and RPA, RDA does not have its own iden-
13
tity and, therefore, acts via the IT infrastructure of its users with the same roles and
access control rights, whereas RPA is working autonomously in the background on
a central server structure (P40). Furthermore, RDA is attended, whereas RPA is
unattended (P40). Attended automation allows the user to monitor the bot as well as
to pause, interrupt or stop it at any time. Furthermore, data can be provided dur-
ing execution. Contrary, unattended automation does not take data during execution.
Once triggered, the unattended bot runs without human involvement. Additionally,
scripting and screen scraping are locally deployed from the user’s desktop and can
be seen as RDA, differing from RPA that meets IT requirements such as security,
scalability, auditability, and changeability (P58). In P51, stand-alone automation in-
cludes macros, office program automation, and mouse/keyboard emulation. Table 4
summarises the differences between RPA and RDA.
Table 4. Differences between RPA and RDA.
Criterion RPA RDA Ref
What? virtual user working on central server personal assistant on user’s desktop P40
How? unattended attended P40
Many publications further distinguish between intelligent and cognitive automation.
Intelligent or enhanced RPA, also denoted as self-learning RPA, uses data to learn
how a user interacts with the system and mimics these interactions including human
judgement (P06, P19, P51). Machine learning and process mining techniques (van der
Aalst 2011) are used to build knowledge of the routines to be automated (P51, P54).
As major advantage, no time-consuming interviews and workshops become necessary;
instead the behaviour of the routine to be automated can be discovered from event logs.
Section 4.4 presents methods for improving RPA projects, including a framework for
combining process mining and RPA (P36). Finally, cognitive RPA relies on advanced
machine learning and natural language processing to augment human intelligence and
to learn how to optimise task performance (P06, P43, P54). The main differences
between rule-based automation and intelligent automation are summarised in Table 5.
Table 5. Differences between RPA and intelligent automation.
Criterion RPA Intelligent Automation Ref
Degree of standardisation high low P58
Data structured unstructured P51, P54
Decisions rule-based knowledge/experience-based P06, P19, P51
Outcome deterministic probabilistic P01, P54
Exceptions demand human intervention trigger machine learning P43, P54
Many publications emphasise the differences between RPA and BPM. Understand-
ing these difference is fundamental for appropriately applying RPA and BPM methods
in an enterprise context. Figure 7 illustrates these differences. The x-axis indicates the
number of process task variants (i.e., the complexity of the process task), whereas the
y-axis displays the case frequency of all process task variants of the business process.
The tasks on the left are suited best for BPM, the ones in the middle are candidate
tasks for RPA, and the ones on the right can be solely performed by humans (cf. Fig-
ure 7) (P56, P60). As can be seen, RPA and BPM complement each other. Enterprises
need to combine the two approaches in the right way to achieve the best automation
results (P60).
Table 6 summarises the main differences between RPA and BPM.
To understand the difference between lightweight and heavyweight IT (Table 6,
14
Figure 4: Comparison of tasks suitable for BPM, for RPA, and tasks only Humans can do.
Case Frequency of Process Tasks
Complexity of Process Tasks
BPM task RPA task Human task
Figure 7. Comparison of tasks suitable for BPM, for RPA, and tasks only Humans can do.
Table 6. Differences between RPA and BPM.
Criterion RPA BPM Ref
Goal automation of selected process
tasks by software bots
automation of business processes
or parts involving the interac-
tions at running time
P07
General idea change ‘where’ work is done change ‘how’ work is done P07, P44, P53
Invasiveness non-invasive, lightweight IT sit-
ting on top of existing business
applications
heavyweight IT interacting with
business logic and creating new
business applications
P01, P07, P46, P60
Problems privacy, security issues high complexity, expensive P44, P46, P53
row 3), we summarise characteristics of suitable tasks for both types of automation.
Lightweight IT automates tasks involving multiple systems and having a high volume.
The systems are characterised by a stable UI. Heavyweight IT, in turn, automates tasks
running in one system, having a very high volume, and relying on a stable back-end
system architecture (P07).
P56 emphasises another differentiation: RPA versus Straight Through Processing
(STP). STP refers to business processes that can be performed without any human
involvement, and, thus, is quite similar to RPA. However, STP has not been widely
adopted in practice as it could only handle few business processes (cf. left side of Fig-
ure 7). RPA is an approach, which uses existing information systems without chang-
ing them, and virtual users interacting with these systems. If the information systems
evolve, the RPA implementation can be adapted as well.
4.2. RQ 2: Which process tasks can be automated with RPA and which
tools are used for automation?
For a successful uptake of RPA initiatives, it is crucial to identify those process tasks
suited for an RPA automation. Discovering corresponding tasks is a challenge on its
own and fosters the automation of business processes or parts of them. This section
provides criteria for selecting process tasks and presents use cases to which RPA has
been successfully applied.
Criteria for selecting process tasks. The most frequently mentioned criterion
in literature is repetitiveness, i.e., the process task to be automated by bots shall
15
have a high volume of transactions or a large number of process task executions (P06,
P10, P14, P28, P31, P50, P58, P59, P62, P63). Regarding the predictability of the
process task volumes, P06 states that process tasks with unpredictable peaks are suited
for RPA implementations. By contrast, P31 emphasises that the volumes should be
predictable.
Another criterion concerns the rule-based character of the process task, i.e., the
process task to be automated shall be standardised, run in a stable environment, and
only require little exception handling (P02, P06, P14, P28, P31, P46, P62, P63).
The next criterion is to check whether the process task requires high manual
efforts and, thus, is prone to errors (P06, P14, P28, P51). Furthermore, digitisation
gaps in business processes might fulfil this criterion as they indicate the need for
human work. P51 even states that ‘any activity that a person performs with mouse
and keyboard can be carried out by a software robot.’
The complexity of the process task itself or, as a result, the complexity of its
implementation, constitutes another selection criterion. All publications agree that
the lower its complexity, the better a process task is suited for RPA (P17, P50, P58,
P59). Moreover, the duration of process task execution can serve as a criterion,
i.e., process tasks to be automated shall have a high expenditure of time (P14, P17).
Additionally, the following criteria are mentioned by a few publications: The inputs
and outputs are digital and structured (P10, P28, P46), the process task only requires a
limited number of human interactions, the process task accesses multiple applications,
the effects of a business failure are high (P14, P28), and the transaction has a significant
influence on the business (P14, P63). P06 proposes choosing process tasks for RPA
automation that do not constitute a priority for the IT department. Figure 8 displays
the six most relevant criteria for selecting process tasks. Note that in related studies,
other numbers of criteria are stated, e.g., 13 in (Santos, Pereira, and Vasconcelos 2019).
However, based on our empirical results the six criteria visualised are the most relevant
ones.
Process
Task
Selection
Criteria
Repetitive
Rule-based
High
manual
efforts
Duration
of task
execution
Digital
inputs and
outputs
Complexity
Figure 8. Criteria for selecting process tasks.
Use Cases. Table 7 summarises the 15 case studies, indicating in which business
area RPA was applied, which process task was automated, and which tool was used
for creating the respective software robot.
The most referred business areas are Business Process Outsourcing (BPO)
(P01, P17, P30), Shared Services (P32, P58), Telecommunication (P50, P59),
and Banking (P40, P53). One case study was conducted in each of the following
areas: Digital Forensics (P03), Auditing (P10), Energy Supply (P31), Manufacturing
(P47), Corporate Service Provider (P61), and Software Testing (P63).
The most frequently automated process tasks in these case studies are swivel-
16
Table 7. Business Area, Process Task, and Automation Tool for concrete Use Cases.
Ref Business Area Process Task Automation Tool
P01 BPO Generate payment receipt -
P03 Digital Forensic Search for keywords within Autopsy forensic soft-
ware and import evidence files, process them and
carry out image extraction in Griffeye forensic
software
UiPath
P10 Auditing Collect data; copy it to template; filter, prepare,
transfer it to database, and perform audit tests
for loan testing
-
P17 BPO Update employee payment details and create new
employment relationships
UiPath
P30 BPO Create and validate Premium Advice Notes Blue Prism
P31 Energy Supply Resolve infeasible customer meter readings -
P32 Shared Services Generate financial close Redwood
P40 Banking Copy details of personal loan or current account
from mainframe application to Excel
Roboplatform
P47 Manufacturing Master data management -
P50 Telecommunication Bundle support tools for field service technician Bluepond
P53 Banking Manage information interaction between bank
and governmental institution
Blue Prism
P58 Shared Services Copy data from Excel to HRM System Blue Prism
P59 Telecommunication Carry out SIM swaps and apply pre-calculated
credit to account
Blue Prism
P61 Corporate Service Provider Generation of documents for annual compliance
process and handle ad-hoc inquiries of customer
-
P63 Software Testing Schedule and control software testing by execut-
ing test scripts and validating the UI
Workfusion
chair processes, i.e., ‘processes where humans take inputs from one set of systems
(for example email), process those inputs using rules, and then enter the outputs into
systems of record (for example Enterprise Resource Planning (ERP) systems)’ (P60).
For ten case studies, the used automation tool was mentioned in the respective
publication: Blue Prism was used in four case studies (P30, P53, P58, P59), UiPath
(P03, P17) in two case studies, and Redwood (P32), Bluepond (P50), Workfusion
(P63), and Roboplatform (P40) in one case study each. Roboplatform is a proprietary
tool that was built by the enterprise itself. P23 compares different automation tools,
namely UiPath, Automation Anywhere, and Blue Prism based on criteria like openness
of the platform, future scope, and performance. Their recommendation is to use UiPath
as it ‘triumphs all’ (P23).
4.3. RQ 3: What are RPA effects?
To decide whether it is worthwhile to adopt RPA, the effects of RPA projects need
to be understood. It is, therefore, crucial to know which RPA effects are reported in
literature, and to exploit this knowledge as benchmark for assessing RPA projects.
Therefore, we merge effects presented in different publications to a holistic view. The
answer to RQ 3 is divided into two aspects: the first one deals with the RPA effects
on humans and their worklife, whereas the second aspect deals with positive, contro-
versially discussed, and negative effects on the company. Figure 9 gives an overview
of the identified RPA effects.
As a positive effect of RPA on employees, the latter are relieved from non-value
adding tasks and, consequently, become more satisfied (P12, P17, P25, P33, P47,
P51, P55, P58). Furthermore, new tasks and jobs for employees are proposed. One
area concerns the development, testing, and monitoring of software robots (P02, P28,
P32). Most publications mention that humans should focus on cognitively more
17
RPA effects
On human
On company
Positive
Negative
Positive
Controversially
discussed
Negative
Relieved from non-value adding process tasks
Focus on cognitively more demanding tasks
Fear to lose the job
Afraid to learn the use of the new technology
Less process tasks for humans
Speed
Availability
Compliance
Quality
Unability to make decisions
Costs
Non-invasiveness
Workaround
Temporary solution
Figure 9. Effects of RPA introduction.
demanding tasks (P13, P17, P32, P41), including tasks that require judgement,
interpretation, and assessment of results (P03, P10, P29, P54, P58). Furthermore,
unstructured tasks (P29, P32, P54), creative tasks (P32), and tasks demanding for
empathy and social interactions (P29, P58) are best suited for humans, e.g., to build
relationships with the customer (P54).
According to (P02, P13, P17, P18, P30, P55), employees are afraid of losing
their job, i.e., they consider the bots as their job competitors (P02, P18) and are
afraid of learning the use of the new technology (P13, P14). Hence, acceptance
problems might arise (P18). P29 and P54 propose combined human bot teams, where
each team member performs the task he or she can do best. In P30, myths about RPA
are demythologised, e.g., ‘RPA is only used to replace humans with technology’. In
turn, P30 is refuted by the fact that more work can be done with the same number of
people and humans are not replaced by technology. According to P61, however, staff
reduction is one effect of RPA implementations.
According to (P14, P32, P54), less tasks will be left for humans after introducing
RPA, especially regarding low-level tasks not requiring any specific qualification. P11
and P12 emphasise that even knowledge workers are affected by lay-off due to RPA.
On one hand, this has an impact on jobs in low-cost countries (P32). P30 proposes
to automate offshore process tasks and keep them offshore, whereas P03 stresses that
humans are needed to trigger the bot. On the other, organisational structures change.
Nowadays, most companies are structured like a pyramid, having many less-skilled
workers and fewer highly skilled workers (cf. Figure 10a). P32 predicts the change from
that pyramid structure to a diamond structure meaning that employees at the bottom
of the pyramid will be replaced by bots (cf. Figure 10b). P42 predicts that the pyramid
structure will be replaced by a pillar structure regarding the human workforce (cf.
Figure 10c). Bots will fill up the structure such that the overall organisation structure
remains a pyramid.
The positive RPA effects on any company, excluding human aspects, can be clus-
tered in four categories:
•Speed: Automated process tasks run faster and task duration becomes shorter
(P01, P12, P14, P18, P28, P29, P35, P47, P50, P54, P55, P58, P61).
18
a b c
Figure 5: Organisational Structure in a Company -Human Workforce is a a: Pyramid b: Diamond c: Pillar.
Figure 10. Organisational Structure in a Company - Human Workforce is a a: Pyramid b: Diamond c: Pillar.
•Availability: Most RPA bots are available 24/7, and instant access is granted.
Moreover, RPA is highly scalable to meet a varying intensity of demands (P01,
P06, P14, P29, P33, P43, P50, P51, P59, P62).
•Compliance: Process tasks executed by a bot are highly transparent and doc-
umented in detail. As a consequence, compliance increases (P29, P32, P33, P35,
P43, P51, P59).
•Quality: RPA eliminates human errors, improves both accuracy and data qual-
ity, and leads to higher customer satisfaction (P06, P12-P14, P28, P35, P43,
P47, P50, P51, P54, P55, P58, P59, P62).
Certain effects of RPA projects are controversially discussed in literature. P43 and
P46 criticise that RPA is unable to make decisions, P50 argues that it provides
full transparency of all decisions. The latter means that if the bot fails, an employee
still can perform the task manually. Section 4.5 discusses how AI is used to expand the
limits of RPA, including decision making. The costs of RPA constitute another point
of discussion: in P14 and P55, budget constraints are seen as a particular challenge for
every RPA project, whereas many publications highlight the cheapness, cost reduction,
and high return on investment coming with RPA (P03, P06, P13, P18, P31-P33, P35,
P46, P47, P50, P54, P59). P18 differs between implementation and maintenance: The
former is characterised by low costs, whereas the latter can be costly and tedious. The
non-invasiveness of RPA is perceived differently: P03 and P46 criticise that RPA
presumes an existing infrastructure and depends on the stability, availability, and per-
formance of the systems. On the other, P54 considers the non-invasiveness of RPA as a
benefit. P04 starts a discussion on possible RPA effects on enterprise architectures and
argues that RPA might become invasive, i.e., RPA enables new workflows, requiring
a modelling functionality in RPA systems, which contradicts the basic RPA idea. P46
emphasises that RPA is unable to adapt to a changing environment, whereas P02 and
P62 notice that RPA implementations are easily modifiable and flexible.
Negative effects or limitations of RPA are seldomly reported. Only P19 characterises
RPA solutions as workarounds, and P02 and P18 point out that they constitute a
temporary solution. According to P03, there are software platforms, e.g., special
forensic software, not compatible with current RPA solutions. Furthermore, P18 crit-
icises that know-how and skills are required, and RPA solutions are not robust in re-
spect to evolving user interfaces. P28 adds that RPA implementations require greater
IT involvement than initially thought.
4.4. RQ 4: Are there methods for improving the implementation of RPA
projects?
To analyse the publications that introduce methods for RPA projects, we oriented
ourselves on the software development life cycle (SDLC) (Royce 1987). We assigned the
19
methods to the corresponding stage in the life cycle for the sake of better illustration
(cf. Figure 11). On one hand, RPA developers asking for support in RPA projects,
can choose an adequate method according to their needs. On the other, researchers
get an overview of existing methods upon which they can build their future work. To
the best of our knowledge, there is no such holistic overview yet. In the following, the
respective methods are described shortly for each development stage.
Analysis
•P05: Analyse UI logs to find deterministic actions
•P16: Process mining to get insights into process tasks
•P22: Re-engineering of tasks
•P25: Monitor screen to discover business process model
•P34: Discovery of data transformations from UI logs
•P35: Process mining to discover business process models
•P37: Tool to record UI logs usable for process mining
•P38: Classify tasks based on textual business process description
•P39: Analyse process tasks to verify suitability for RPA
•P48: Evaluate process tasks with multiple criteria
•P57: Decision support to prioritise process tasks while
maximising RPA benefits
Product Design
•P44: Organisation of RPA in local business units
Coding
•P07: Agile development with test driven development
•P27: RPA development using a digital twin
Testing
•P08: Automatic testing
•P24: Automatic generation of testing environment
•P35: Automatic training of bots
Operation
•P16: Process mining for monitoring
•P21: Job-scheduling algorithm for task assignment to RPA bots
•P52: Calculate optimal number of software licenses and task
assignment
Analysis, Product
Design, and Coding
•P15: RPA rule deduction from user
behaviour
•P36: Robotic Process Mining pipeline
to generate RPA scripts from UI
logs
Product Design, Coding, and
Testing
•P49: Framework to transform human-
centred routine into robot-automated
routine using guidelines
Complete Life Cycle
•P20: Framework for applying
RPA in auditing
Figure 6: SDLC annotated with Method Paper.
Figure 11. SDLC annotated with Publications providing Methods for RPA Projects.
Analysis Stage. The approaches to improve the Analysis Stage can be clustered
into three areas: business process insights,business process standardisation,
and process task selection.
P16 uses process mining to get insights into the business process, e.g., its automa-
tion rate. In P38, textual business process descriptions are used to classify the tasks
into the categories Manual Task,User Task, and Automated Task. The goal is to auto-
matically detect process tasks suited for RPA. To achieve this goal, P38 uses feature
computation for prediction and a Support Vector Machine for classifying the business
process descriptions based on the features.
The aim of P35 is to develop a new process mining technique that can deal with
RPA and automatically discover business process models. The approach is to dis-
cover constraints within an event log, extract corresponding feature vectors, and label
constraint violations. P35 uses clustering methods to identify correlations between ac-
tivation and target payloads. In a subsequent publication of the same authors, i.e.
P37, a tool (‘Action Logger’) is developed that records UI logs, which can directly
serve as input to process mining tools. The UI logs contain all information relevant for
RPA implementation. Finally, P34, again contributed by the same authors, develops
an idea how to discover data transformations from UI logs.
In P16, process mining is used to standardise existing business processes. Another
20
standardisation technique is proposed by P22, which emphasises the importance of
not automating the as-is business process, but to optimise it before automation. Thus,
the authors propose a framework for business process re-engineering.
The most difficult task in the analysis stage is to select the process tasks to be au-
tomated by a bot. Different approaches are proposed: P16 sticks to process mining for
prioritising process tasks. P25 also uses process mining to discover process tasks, with
a method focusing on creating event logs from screen monitoring data. P05 analyses
UI logs to discover deterministic actions. As basic idea, ‘a routine is automatable if
its first action is always triggered when a condition is met [...] and the value of each
parameter of each action can be computed from the values of parameters of previous
actions’ (P05).
P39 develops a four-step method to analyse a process task based on its character-
istics (cf. Section 4.2): first, to be eligible for RPA, the process task has to be mature
and standardised (Step 1). Step 2 assesses the RPA potential of the process task based
on human interaction with the information systems and its rule-based nature. Step 3
evaluates the RPA relevance according to the volume of transactions and the degree of
process task complexity. Finally, based on Steps 2 and 3, the process task is classified.
P39 recommends selecting process tasks with high relevance and high potential. P48
follows a similar approach and develops a multi-criteria process task evaluation model,
which assesses the technical feasibility and business potential criteria to find suitable
process tasks for RPA. The technical criteria include the degree of rule-basedness,
human intervention, digitalisation, and the structuredness of data. The potential cri-
teria evaluate labour intensity, the number of systems involved, the number of process
task exceptions, the number of task steps, current costs, and process task maturity.
P57 proposes a method to prioritise process tasks while maximising RPA benefits.
Based on different indicators of the process task, i.e., execution frequency, execution
time, degree of standardisation, stability (i.e., small number of exceptions in the pro-
cess task), failure rate, and automation rate, the automation potential of the process
task is assessed. Furthermore, the profitability of process task automation is measured
through fixed and variable costs of human labour as well as fixed and variable costs
of RPA. Finally, P57 maximises the economic value and provides recommendations to
support the decision of selecting appropriate processes for RPA initiatives.
Product Design Stage. P44 highlights advantages and challenges of organising
RPA in local business units. On the one hand, enthusiasm for digitalisation and local
ownership are built. On the other, there is a lack of control mechanisms and end-to-
end business process views. P44 proposes to loosely couple the IT department and the
RPA team.
Coding Stage. P07 suggests a method for implementing RPA projects in an agile
way: instead of documenting a process task completely with clicks and text-based
description, the users themselves record a video when performing the process task and
store the latter in the backlog. The developers create a test case for this video and
checks whether the current solution passes the test (Test Driven Development). If not,
they modify the RPA solution until the test case is fulfilled. Then, they move on to
the next video. P27 proposes the use of digital twins for RPA development. In this
context, a digital twin corresponds to a virtual shadow of an IT system. The idea
allows developing RPA externally without having access to the real system.
Testing Stage. P35 has the vision to develop a method ‘to automatically train
the RPA bots’. Research has not progressed far enough. P08 proposes a method for
automated testing in RPA projects, which has been validated with a prototype. The
approach is to modify the RPA life cycle. Compared to the life cycle model depicted
21
in Figure 11, the third stage is called development and not coding, operation is named
monitoring, and a fourth stage, i.e., deployment, is inserted before the testing phase.
The modified life cycle not only includes design in the second stage, but test environ-
ment construction as well. During development, automatic testing can be performed
serving as new input for the analysis phase. P24 extends the idea of P08 by providing
technical details on test cases and the algorithm as well as by evaluating the approach
of automatically generating a testing environment.
Operation Stage. P16 mentions that process mining can be used to monitor
the results of an RPA project. P21 proposes a middleware system for controlling
the execution of multiple RPA bots. The system includes a job-scheduling algorithm
to efficiently distribute multiple tasks among available bots. In turn, P52 solves
an optimisation problem to determine the optimal number of required bots while
minimising costs. Then, the optimal task assignment among the bots is solved.
Some publications cannot be assigned to solely one stage and are, therefore, placed in
the centre of Figure 11. P15 and P36 cover the first three stages, i.e., Analysis, Product
Design, and Coding. To be more precise, P15 presents an end-to-end approach that
allows deducing RPA rules from user behaviour. The idea is based on the Form-to-
Rule approach: First, tasks of the user are identified by observing interactions with
systems and identifying forms used within the systems. Second, rules are deduced
from relations between the different tasks. Third, RPA is implemented based on those
rules. P36 combines the approaches presented in P34, P35, and P37 and proposes a
Robotic Process Mining pipeline (cf. Figure 12). After recording UI logs, noise filtering,
segmentation, and simplification steps are applied to identify candidate routines. In
these routines, executable (sub)routines are discovered and compiled to obtain RPA
scripts. P36 emphasises that there are still many challenges to successfully apply the
proposed pipeline. However, as main advantage, the pipeline presents a first end-to-
end automation. Therefore, it serves as a reference for other approaches with process
mining.
Recording
UI Log
Noise filtering
Executable
(sub)routine
discovery
Candidate
routine
identification
SimplificationSegmentation Compilation
RPA Scripts
Executable
(sub)routine
specifications
Candidate
routines for
automation
Simplified
Task Traces
Task Traces
Cleaned
UI Log
Steps:
Outcome:
Figure 12. Robotic Process Mining Pipeline according to (Leno et al. 2020b).
Stages Product Design, Coding, and Testing are addressed by P49. A framework
is developed to transform a human-centred routine into a robot-automated one. The
framework of routine automation can be empirically applied to different areas, includ-
ing RPA, and provides implementation guidelines.
One publication, i.e., P20, addresses the complete life cycle of RPA and proposes a
framework for introducing RPA in auditing. The first stage is process task selection
based on the evaluation of different criteria, e.g., RPA criteria (cf. Section 4.2), process
task complexity, and compatibility of the data used in the task automated with RPA.
Second, the process task is modified, e.g., considering data standardisation, i.e., to
unify data from multiple sources. In a third step, the process task is implemented
and, finally, evaluated and operated. The last step consists of evaluating effectiveness,
assessing detection risk (i.e. the risk that auditing ‘will not detect a misstatement’
(P20)), and monitoring the RPA operations.
22
4.5. RQ 5: Is AI used in combination with RPA?
Due to the growing importance of AI in automation, RQ 5 investigates whether AI is
used in combination with RPA and - if yes - how RPA can profit from AI. To get an
idea whether AI has already been used in combination with RPA, our insights from
literature are summarised in the following.
Some works briefly mention the use of AI and its potentials. P40 and P43 state
that with AI it becomes possible to understand semi-structured data. This can be
used to extract data relevant for the RPA bot from unstructured data and to relax the
restriction on structured input data (cf. Section 4.2). P56, in turn, emphasises that
AI helps to interpret changing user interfaces and to improve the robustness of RPA
solutions. Using chat bots, P43 presumes that the interaction between humans and
computer systems becomes facilitated.
First AI-based applications have emerged in the RPA field: P26 presents a Cog-
nitive Automation Robots Platform, which is able to understand data, generate in-
sights, and use the latter as learning experiences. P33 uses the cognitive virtual agent
‘Amelia’, which understands chat messages. In P19, a cognitive RPA prototype is
presented, which can automatically identify, extract, and process data. Once the clas-
sification model is trained (for details see P19, pp. 68-69), new unseen documents are
classified and relevant objects, e.g., address fields, are detected and extracted.
We discovered four publications that combine AI with RPA in greater detail. P41
provides building blocks for intelligent process automation by explaining and provid-
ing implementations on how to extract intentions from audio, classify emails, detect
anomalies, find cross correlations in time series, and understand traffic patterns. P62
describes how machine learning methods contribute to further improve RPA, e.g., us-
ing image processing to scan letters or invoices or using classification algorithms to
label documents. The task of classifying emails correctly is picked up by several
publications: P45 proposes the use of an Support Vector Machine and a Text Rank
Algorithm to read emails and to automatically process them. P09 develops an algo-
rithm, named Sure-Tree, for email classification, which produces a minimum of false
positives to ensure that an incorrect action is never triggered.
5. Deriving a Framework for Analysing and Comparing RPA
Publications
This section synthesises the results obtained by the SMS. More precisely, we present a
classification framework for Analysing and Comparing existing as well as upcoming
Publications in RPA (ANCOPUR for short). ANCOPUR gathers the results along the
derived research questions (cf. Section 3.1). The section presents the framework and
explains it with an example (cf. Section 5.1). Furthermore, researchers in a large au-
tomotive enterprise use the framework in their daily work. We illustrate its usefulness
and practical relevance with this practical application (cf. Section 5.2).
5.1. Explanation of ANCOPUR
Tables 8 and 9 depict the schema of the ANCOPUR framework we derived: The first
column shows the main aspect for comparison, e.g., meaning, criteria for selecting
process tasks, or use case. In the following columns the aspect gets detailed. The
publication can be assigned to several rows depending on the aspects it covers. If a
23
new feature is found, it can be added to ANCOPUR as well, i.e., we consider our
classification and evaluation framework as extensible. To demonstrate its usefulness
and applicability, all 63 publications from the SMS are categorised with ANCOPUR.
Note that this significantly facilitates the comparison of any new publication with
existing knowledge as well. We illustrate and explain ANCOPUR by assigning P17
exemplarily to it. This publication was one of the 63 identified as relevant by the
SMS.
Publication P17 is read to detect information depositing on the first column of the
ANCOPUR framework. We discover the aspects criteria for selecting process tasks,
Effects, and Use Case. Concerning the criteria for process tasks to be automated, P17
emphasises that ‘1) the processes should be simple enough so that the robots could
be implemented quickly and 2) improved process efficiency resulting from RPA imple-
mentation should be clearly visible’ (Hallikainen, Bekkhus, and Pan 2018). Therefore,
the process tasks are selected depending on their complexity and the duration of
process task execution. Applying ANCOPUR, P17 is indeed assigned to those two
rows in the criteria for selecting process tasks column (cf. Table 8). Regarding the
case study, the aspects Business Area,Business Process Task, and Automation Tool
are all covered by P17: the general business area is BPO and the concrete processes
are ‘1) new employment relationships and 2) changes in employee payment details’,
which are both swivel-chair processes. UiPath was used for automating the business
processes. Therefore, P17 is added as reference to rows BPO,swivel-chair process,
and UiPath in the use case section of ANCOPUR (cf. Table 8). The following wording
is found for RPA effects: ‘there were some fears about losing jobs [...] people would
no longer have to carry out the boring work and could concentrate on more inter-
esting tasks’ (P17). The first statement expresses a negative effect on humans and
is assigned to fear to lose the job in ANCOPUR. The second statement describes
positive effects on humans and covers both aspects in ANCOPUR, namely relieved
from non-value adding tasks and focus on cognitively more demanding tasks
(cf. Table 9).
The assignment of P17 to ANCOPUR eases its systematic comparisons with other
publications. Assume that one is interested in RPA Case Studies and has read P17
and classified it with the framework. There is another publication using the same
automation tool, i.e., UiPath, namely, P03. Hence, one can read P03 and compare it
with P17. Further, one becomes aware of other automation tools. Concerning RPA
Effects, P17 only covers effects on humans. Here, one can find further publications
covering the same (e.g., P02, P55) or different aspects (e.g., P14). Additionally, effects
on the company are not covered in P17, therefore, with ANCOPUR one can determine
relevant publications covering this aspect.
5.2. ANCOPUR Application in Practice
The classification framework is initially presented to researchers in a large automo-
tive company to assess its applicability and complexity. The researchers are asked to
assign a publication randomly excluded in the results of the SMS to the framework
(Flechsig, Lohmer, and Lasch 2019). For Meaning,Use Case,Effect, and Combination
with AI, no new aspects or no information are found at all. Concerning Differences of
RPA to Related Technologies,BPM is compared to RPA. All aspects in ANCOPUR
are covered, only the formulations differ a bit, e.g., ‘Redesign of extensive processes
with high strategic relevance and added value’ (Flechsig, Lohmer, and Lasch 2019) is
24
assigned to the row changes ‘how’ work is done. Regarding criteria for selecting
process tasks, the researchers find several questions that aim to identify processes suit-
able for an RPA implementation (Flechsig, Lohmer, and Lasch 2019, p. 111). These
criteria include repetitiveness,rule-based,duration of process task execution,
and high effects of business failure. Therefore, (Flechsig, Lohmer, and Lasch 2019)
may be added to the corresponding rows in the ANCOPUR framework. Additionally,
(Flechsig, Lohmer, and Lasch 2019) proposes choosing processes relevant for com-
pliance, an aspect not considered yet. The researchers discuss and decide to expand
ANCOPUR by this process task selection criterion aspect. (Flechsig, Lohmer, and
Lasch 2019) suggests a method for combining BPM and RPA, which can be assigned
to the Product Design Stage with a new row, namely combination of BPM and
RPA. The idea is to have a common Analysis Stage for BPM and RPA projects as
well as to decide in the Product Design Stage whether to implement a BPM or an RPA
solution. Overall, the researchers agree that ANCOPUR provides a useful overview for
RPA research and is easy to use and, especially, to expand.
After this initial positive contact with the framework, we spread ANCOPUR
throughout the enterprise. Since then, it has been helpful in various situations:
•Colleagues from the IT department investigated which task selection criteria can
be found in literature to adapt the identification of tasks suited for automation
to the state-of-the-art. Relevant publications were selected from ANCOPUR.
•Employees from different business areas, e.g., banking asked to what extent RPA
has been used successfully in this area. Thanks to ANCOPUR, the request could
be answered and specific publications could be named.
•Managers are always interested in what benefits automation has on the enter-
prise. With ANCOPUR, known RPA effects can be easily given.
•The works council was interested in which effects on people have to be considered
in RPA projects. We were able to provide specific information here.
Therefore, accompanying different RPA initiatives over a longer time, we recognise
that the framework is highly appreciated and used. The people working with it are
convinced of its advantages and recommend ANCOPUR to their colleagues.
Altogether, ANCOPUR uses criteria and sub-criteria to classify RPA publications.
The framework is useful for systematically analysing, assessing, and comparing existing
as well as upcoming RPA works.
6. Discussion
The main motivation of conducting this SMS was to provide a holistic RPA overview
(cf. Section 1.1) covering different disciplines. The obtained results enable us to not
only answer the five research questions (cf. Section 3.1), but also to derive a classifi-
cation and mapping framework that is useful for both academics and practitioners. In
the following, the results are discussed and interpreted along the seven discovered the-
matic clusters (cf. Section 4). The role of RPA is to automate selected process tasks
or parts, and thus constitutes a crucial tool fostering business process optimisation
due to the automation of individual tasks. The paper explains for which tasks RPA is
best suited (cf. Section 4.2), how RPA projects can be carried out (cf. Section 4.4), and
what effects can be achieved with RPA (cf. Section 4.3). Altogether, this has enabled
a profound positioning of the RPA technology.
More precisely, we identified, categorised, and analysed 63 publications belonging to
25
Table 8. Classification Framework ANCOPUR.
ANCOPUR Ref
Meaning
software-based solution (IEEE 2017), P02, P33, P40, P43, P51, P58, P59, P60
mimics human behaviour P06, P40, P43, P45, P51, P53, P56, P59, P60
characteristics of the
automated process task
rule-based (IEEE 2017), P01, P10, P29, P45
structured data P01, P10, P29, P53
routine tasks P01, P03, P06, P10, P51
non-invasive (IEEE 2017), P06, P46, P51
goal of implementation deliver result/service (IEEE 2017)
humans handle exceptions (IEEE 2017)
Differences
of RPA to
Related
Technology
RDA no own identity P40
attended automation P40
Intelligent/Cognitive
Automation
low degree of standardisation P58
unstructured data P51, P54
knowledge/experience-based decisions P06, P19, P51
probabilistic outcome P01, P54
exceptions trigger machine learning P43, P54
BPM
re-engineer process tasks to optimise them P07
changes ‘how’ work is done P07, P44, P53
creates new business applications P01, P07, P46, P60
highly complex and expensive P44, P46, P53
STP P56
Criteria for
selecting
process
tasks
repetitive P06, P10, P14, P28, P31, P50, P58, P59, P62, P63
rule-based P02, P06, P14, P28, P31, P46, P62, P63
high manual effort P06, P14, P28, P51
complexity P17, P50, P58, P59
duration of process task execution P14, P17
digital and structured inputs and outputs P10, P28, P46
limited number of human intervention P14
access to multiple applications P14, P28
high effects of business failure P14, P63
Use Case
Business Area
BPO P01, P17, P30
Shared Services P32, P58
Telecommunication P50, P59
Banking P40, P53
Digital Forensics P03
Auditing P10
Energy Supplier P31
Manufacturing P47
Corporate Service Provider P61
Software Testing P63
Process Task swivel-chair process P01, P03, P10, P17, P30-P32, P40, P47, P50, P53,
P58, P59, P61, P63
Automation Tool
Blue Prism P30, P53, P58, P59
UiPath P03, P17
Redwood P32
Bluepond P50
Workfusion P63
Roboplatform P40
26
Table 9. Framework ANCOPUR - Continuation.
Effect
on humans and future
worklife
positive relieved from non-value adding tasks P17, P25, P33, P47, P51, P55, P58
focus on cognitively more demanding tasks P02, P03, P10, P12, P13, P17, P28, P29, P32, P41,
P54, P58
negative
fear to lose the job P02, P13, P17, P18, P30, P55
afraid to learn new technology, acceptance problems P13, P14, P18
less tasks, lay-off P11, P12, P14, P32, P42, P54, P61
on company
positive
speed P01, P12, P14, P18, P28, P29, P35, P47, P50, P54,
P55, P58, P61
availability P01, P06, P14, P29, P33, P43, P50, P51, P59, P62
compliance P29, P32, P33, P35, P43, P51, P59
quality P06, P12-P14, P28, P35, P43, P47, P50, P51, P54,
P55, P58, P59, P62
controversially
discussed
inability to make decisions P43, P46, P50
costs P03, P06, P13, P14, P18, P31-P33, P35, P46, P47,
P50, P54, P55, P59
non-invasiveness P03, P46, P54
negative workaround, temporary solution P02, P18, P19
incompatibility of software with RPA solutions P03
Method
Analysis Stage
process task in-
sights
automation rate P16
classify tasks based on textual business process description P38
discover business process models P25, P34, P35, P37
process task
standardisation framework for process task re-engineering P22
process task se-
lection
multi-criteria process task evaluation model P48
analyse process tasks based on criteria P39
prioritise process tasks, maximise benefits P57
analyse UI logs to find deterministic actions P05
Product Design Stage organise RPA in local business units P44
Coding Stage agile development P07
development with digital twin P27
Testing Stage automatic testing P08, P24
automatic training P35
Operation Stage
process mining to monitor results P16
algorithm for job-scheduling and task assignment P21
optimal number of licences and task assignment P52
Several Stages
RPA rule deduction from user behaviour P15
generate RPA scripts from UI logs P36
framework to transform human-centred into robot-
automated routine P49
framework for RPA in auditing P20
Combination
with AI
briefly mention AI P40, P43, P56
present prototype P19, P26, P33
machine learning methods P41, P62
classify emails Support Vector Machine, Text Rank P45
Sure-Tree P09
27
the following seven clusters: RPA Meaning, Differences of RPA to Related Technolo-
gies, Criteria for selecting process tasks, RPA Use Cases, RPA Effects, RPA Project
Methods, and Combination of RPA with AI. As main result we obtain the ANCOPUR
framework, which enables a structured overview of existing work on RPA as well as the
identification of research gaps. More specifically, the ANCOPUR classification frame-
work provides a fast and easy way to identify and categorise publications in the RPA
area. Hence, the main goal of conducting an SMS is fulfilled (Petersen et al. 2008). Note
that some publications cover several components of the framework, whereas others fo-
cus on one specific aspect. For example, P08 presents a method for testing, no other
aspect is covered. In P29, the meanings of RPA and effects of RPA projects are dis-
cussed. However, the focus is on effects on the company (organisational/management
aspect). P59 covers nearly all aspects, i.e., RPA is defined, criteria for selecting pro-
cess tasks are provided, the tool used for automation is explained, and effects on the
company resulting from the RPA project are presented. ANCOPUR merges the differ-
ent aspects of all publications and provides a holistic overview on RPA literature. In
particular, comparing new works with existing knowledge becomes much simpler and
more structured. Moreover, ANCOPUR can be easily expanded. If new publications
reveal unconsidered aspects, those can be added to evolve the framework and keep it
up to date. The added value of ANCOPUR has been confirmed by engineers in a large
automotive company working with the classification framework. In detail:
1. RPA Meanings. It is emphasised that RPA is a software-based solution mim-
icking human behaviour. These aspects are crucial for indicating the difference
of RPA to hardware bots.
2. Differences of RPA to Related Technologies. Most papers emphasise the
differences between RPA and Intelligent Automation as well as between RPA
and BPM.
3. Criteria for selecting process tasks. Best suited for an RPA automation are
repetitive, rule-based, and complex process tasks demanding for high manual
efforts.
4. Use Cases. The majority of use cases stem from business areas such as BPO
and Shared Services. Note that this is reasonable as those areas possess many
repetitive, rule-based process tasks as, for example, generation of payment re-
ceipt (Aguirre and Rodriguez 2017). Anyway, it would be interesting to encounter
more RPA projects in knowledge-intensive business areas, e.g., in the research
and development field or in healthcare. Furthermore, current literature only re-
ports on successful RPA projects, leaving room for further research on failed
projects. Concerning the RPA tools used in the case studies, Blue Prism and
UiPath are dominating. According to (Gartner 2019), however, there are other
tools that should be considered: Automation Anywhere, EdgeVerve Systems,
NICE, Workfusion, Pegasystems, and Another Monday. The application of the
different tools to one concrete use case as well as a systematic comparison of tool
performance should be subject of further research studies.
5. RPA Effects. The positive effects of RPA are widely discussed in literature.
Only a minority is critical towards RPA. A potential reason for this is the novelty
of RPA (cf. Figure 5 in Section 4), due to which the technology is hyped and
negative effects do not want to be seen. It is emphasised that employees are
relieved from non-value adding tasks, and instead may focus on cognitively more
demanding tasks. Finally, process tasks become faster, better available, more
compliant, and improved in quality.
28
6. RPA Project Methods. Most methods improving RPA implementation were
published in 2019 and 2020 (16 of 22 papers). The vast majority of methods
tries to improve the analysis stage, only some publications address the other life
cycle stages. The analysis stage is the one that differs mostly from other software
development projects. Product design, coding, and testing are not differing that
much when either implementing an RPA project or any other software project.
We expect that more publications adding analysis issues will appear as well as
methods to fully automate the discovery of RPA-suitable process tasks. Fur-
thermore, the operation stage should be addressed, e.g., it should be monitored
whether the bots are accepted or employees fear to lose their job and, therefore,
refuse the use of the bots.
7. Combination of RPA with AI. The use of AI in the context of RPA is still at
a rather early stage. Six publications deal with this combination from a general
point of view and emphasise that it might create a big impact. Only four concrete
use cases are discovered, the majority focuses on the problem of classifying emails
correctly. While the use cases are still scientific in nature, it will be interesting
to learn more about industry-driven approaches and projects. The publications
are from the last years only, therefore, we hope for more research in the coming
years.
In general, research on RPA is still at its beginning. Though being increasingly
present in industry, scientific works on this topic are rather scarce and mainly consider
qualitative issues. Moreover, it is noteworthy that quantitative research is missing. We
expect that there will be a lot more publications in the coming years. In order to assess
and compare those publications with the existing body of knowledge, the present paper
provides a fundamental framework based on concepts of RPA.
7. Summary and Outlook
RPA is a novel technology that emerged in 2015. By means of an SMS, we provide an
overview of the most relevant publications until June 2020. We discover seven thematic
clusters answering fundamental questions such as ‘What meaning is attached to RPA
in literature?’, ‘Which process tasks can be automated with RPA?’, and ‘What are the
RPA effects?’. Furthermore, we investigate the differences between RPA and related
technologies, methods for improving the implementation of RPA projects, and whether
AI is used in combination with RPA. Additionally, we provide a review of case studies
including the business area, process task, and the automation tool.
The SMS results in ANCOPUR, a framework for systematically analysing and
comparing emerging publications in the RPA area. With the help of criteria and
sub-criteria, publications can be classified. The framework provides a robust and ex-
pandable systematics to categorise and evaluate trends and further developments in
the RPA area. It is already in use in a large automotive company for these reasons.
Therefore, it will help both scientists and users from industry to assess and compare
upcoming RPA publications.
As discussed, due to the novelty of RPA, the research focus lies on analysing and
understanding the RPA technology. The Combination of AI with RPA and the devel-
opment of Methods for RPA implementation are still in their beginning. Regarding the
publication dates of the respective publications, there is a clear trend in this direction
visible: nine of ten publications combining RPA and AI and 20 of 22 method papers
29
were published in 2018, 2019, and 2020.
References
Aguirre, Santiago, and Alejandro Rodriguez. 2017. “Automation of a Business Process Using
Robotic Process Automation (RPA): A Case Study.” In Workshop on Engineering Appli-
cations, 65–71. Springer, Cham.
Asatiani, Aleksandre, and Esko Penttinen. 2016. “Turning robotic process automation into
commercial success - Case OpusCapita.” Journal of Information Technology Teaching Cases
6 (2): 67–74.
Asquith, Alisha, and Graeme Horsman. 2019. “Let the robots do it! - Taking a look at Robotic
Process Automation and its potential application in digital forensics.” Forensic Science
International: Reports 1.
Auth, Gunnar, and Frank Bensberg. 2019. “Impact of Robotic Process Automation on En-
terprise Architectures.” In INFORMATIK 2019: 50 Jahre Gesellschaft f¨ur Informatik -
Informatik f¨ur Gesellschaft (Workshop Beitr¨age), Gesellschaft f¨ur Informatik eV.
Bosco, Antonio, Adriano Augusto, Marlon Dumas, and Marcello La Rosa. 2019. “Discover-
ing Automatable Routines From User Interaction Logs.” In International Conference on
Business Process Management, 144–162. Springer, Cham.
Bruno, Jerry, Sam Johnson, and Josh Hesley. 2017. “Robotic disruption and the new revenue
cycle.” Healthcare Financial Management 71 (9): 54–62.
Cewe, Christoph, Daniel Koch, and Robert Mertens. 2017. “Minimal Effort Requirements Engi-
neering for Robotic Process Automation with Test Driven Development and Screen Record-
ing.” In International Conference on Business Process Management, 642–648. Springer,
Cham.
Chac´on-Montero, J., A. Jim´enez-Ram´ırez, and J. G. Enr´ıquez. 2019. “Towards a Method
for Automated Testing in Robotic Process Automation Projects.” In 14th International
Workshop on Automation of Software Test, 42–47. IEEE Press.
Chalmers, Eric. 2018. “Machine Learning with Certainty: A Requirement for Intelligent Process
Automation.” In 17th IEEE International Conference on Machine Learning and Applica-
tions, ICMLA, 299–304. IEEE.
Cohen, Michael, Andrea M. Rozario, and Chanyuan Abigail Zhang. 2019. “Exploring the Use
of Robotic Process Automation (RPA) in Substantive Audit Procedures.” The CPA Journal
89 (7): 49–53.
Eikebrokk, Tom, and Dag Olsen. 2019. “Robotic Process Automation for Knowledge Workers
- Will It Lead To Empowerment or Lay-Offs?” In Norsk konferanse for organisasjoners bruk
at IT, Vol. 27, 2015–2017.
Eikebrokk, Tom Roar, and Dag H˚akon Olsen. 2020. “Robotic Process Automation and Con-
sequences for Knowledge Workers; a Mixed-Method Study.” In Conference on e-Business,
e-Services and e-Society, 114–125. Springer.
El-Khawaga, Ghada, Mervat Abuelkheir, Sherif I Barakat, Alaa M Riad, and Manfred Reichert.
2020. “CONDA-PM—A Systematic Review and Framework for Concept Drift Analysis in
Process Mining.” Algorithms 13 (7): 161.
Enriquez, J. G., A. Jimenez-Ramirez, F. J. Dominguez-Mayo, and J. A. Garcia-Garcia. 2020.
“Robotic Process Automation: A Scientific and Industrial Systematic Mapping Study.”
IEEE Access 8: 39113–39129.
Fernandez, Dahlia, and Aini Aman. 2018. “Impacts of Robotic Process Automation on Global
Accounting Services.” Asian Journal of Accounting and Governance 9: 123–132.
Flechsig, Christian, Jacob Lohmer, and Rainer Lasch. 2019. “Realizing the Full Potential of
Robotic Process Automation Through a Combination with BPM.” In Logistics Management,
Lecture Notes in Logistics, edited by C. Bierwirth, T. Kirschstein, and D. Sackmann, 104–
119. Springer, Cham.
30
Fung, Han Ping. 2014. “Criteria, Use Cases and Effects of Information Technology Process
Automation (ITPA).” Advances in Robotics & Automation 3 (3).
Gao, Junxiong, Sebastiaan J. van Zelst, Xixi Lu, and Will M.P. van der Aalst. 2019. “Auto-
mated robotic process automation: A self-learning approach.” Lecture Notes in Computer
Science 11877 LNCS: 95–112.
Gartner. 2019. Magic Quadrant for Robotic Process Automation. Technical Report. Gartner.
Geyer-Klingeberg, Jerome, Janina Nakladal, and Fabian Baldauf. 2018. “Process Mining and
Robotic Process Automation: A Perfect Match.” In BPM (Dissertation/Demos/Industry),
124–131.
Gotthardt, Max, Dan Koivulaakso, Okyanus Paksoy, Cornelius Saramo, Minna Martikainen,
and Othmar M. Lehner. 2019. “Current State and Challenges in the Implementation of
Robotic Process Automation and Artificial Intelligence in Accounting and Auditing.” Jour-
nal of Finance & Risk Perspectives 8: 31–46.
Hallikainen, Petri, Riitta Bekkhus, and Shan L Pan. 2018. “How OpusCapita Used Internal
RPA Capabilities to Offer Services to Clients.” MIS Quarterly Executive 17 (1): 41–52.
Hindel, Julia, Lena M Cabrera, and Matthias Stierle. 2020. “Robotic Process Automation:
Hype or Hope?” In 15th International Conference on Wirtschaftsinformatik, .
Houy, Constantin, Maarten Hamberg, and Peter Fettke. 2019. “Robotic Process Automation in
Public Administrations.” In Digitalisierung von Staat und Verwaltung, edited by M. R¨ackers,
S. Halsbenning, D. R¨atz, D. Richter, and E. Schweighofer, 62–74. Gesellschaft f¨ur Informatik
eV.
Huang, Feiqi, and Miklos A. Vasarhelyi. 2019. “Applying robotic process automation (RPA)
in auditing: A framework.” International Journal of Accounting Information Systems 35.
Hwang, Myeong Ha, Ui Kyun Na, Seungjun Lee, Byung Joo Cho, Yeri Kim, Dong Hyuk Lee,
and Ji Kang Shin. 2020. “MIORPA: Middleware System for open-source robotic process
automation.” Journal of Computing Science and Engineering 14 (1): 19–25.
Ian, Herbert, Aravindhan Dhayalan, and Scott Andy. 2016. “The future of professional work:
Will you be replaced or will you be sitting next to a robot?” Management Services Journal
60 (2): 22–27.
IEEE. 2017. “IEEE Guide to Terms and Concepts in Intelligent Process Automation.” IEEE
Std 2755-2017 1–16.
Issac, Ruchi, Riya Muni, and Kenali Desai. 2018. “Delineated Analysis of Robotic Process
Automation Tools.” In 2nd International Conference on Advances in Electronics, Computers
and Communications, 1–5. IEEE.
Ivanˇci´c, Lucija, Dalia Suˇsa Vugec, and Vesna Bosilj Vukˇsi´c. 2019. “Robotic Process Automa-
tion: Systematic Literature Review.” In Business Process Management: Blockchain and
Central and Eastern Europe Forum, BPM 2019, Lecture Notes in Business Information
Processing, edited by C. et al. Di Ciccio, Vol. 361, 280–295. Springer, Cham.
Jalali, Samireh, and Claes Wohlin. 2012. “Systematic literature studies: database searches vs.
backward snowballing.” In Proceedings of the 2012 ACM-IEEE international symposium on
empirical software engineering and measurement, 29–38.
Jim´enez-Ram´ırez, A., Hajo Reijers, Irene Barba, and Carmelo Del Valle. 2019. “A Method to
Improve the Early Stages of the Robotic Process Automation Lifecycle.” In International
Conference on Advanced Information Systems Engineering, 446–461. Springer, Cham.
Jim´enez-Ram´ırez, Andres, Jes´us Chac´on-Montero, Tomasz Wojdynsky, and Jos´e Gonz´alez
Enr´ıquez. 2020. “Automated testing in robotic process automation projects.” Journal of
Software: Evolution and Process 1–11.
Khanra, Sayantan, Amandeep Dhir, AKM Najmul Islam, and Matti M¨antym¨aki. 2020. “Big
data analytics in healthcare: a systematic literature review.” Enterprise Information Sys-
tems 14 (7): 878–912.
Kin, Goon Wooi, Rosnin Mustaffa, M.A.J. Johari, and A.Y. Amruddin. 2018. “Cognitive
Automation Robots (CAR).” In E-Proceeding of the 7th International Conference on Social
Science Research, 1–11.
Kitchenham, B., and S. Charters. 2007. Guidelines for performing systematic literature reviews
31
in software engineering. Technical Report. Technical report, EBSE Technical Report. EBSE-
2007-01.
Koch, Julian, Michael Trampler, Ingo Kregel, and Andr´e Coners. 2020. “‘Mirror, Mirror, on
the Wall’: Robotic Process Automation in the Public Sector using a Digital Twin.” In 28th
European Conference on Information Systems (ECIS2020), .
Kokina, Julia, and Shay Blanchette. 2019. “Early evidence of digital labor in accounting: Inno-
vation with Robotic Process Automation.” International Journal of Accounting Information
Systems 35.
Lacity, Mary, and Leslie Willcocks. 2017. “A New Approach to Automating Services.” MIT
Sloan Management Review .
Lacity, Mary, Leslie Willcocks, and Andrew Craig. 2015a. “Robotic process automation at
Xchanging.” The Outsourcing Unit Working Research Paper Series 15 (3): 1–26.
Lacity, Mary, Leslie Willcocks, and Andrew Craig. 2015b. “Robotic Process Automation: Ma-
ture Capabilities in the Energy Sector.” The Outsourcing Unit Working Research Paper
Series 1–19.
Lacity, Mary, Leslie Willcocks, and Andrew Craig. 2016. “Robotizing Global Financial Shared
Services at Royal DSM.” The Outsourcing Unit Working Research Paper Series 16 (2):
1–26.
Lacity, Mary, Leslie Willcocks, and Andrew Craig. 2017. “Service Automation: Cognitive Vir-
tual Agents at SEB Bank.” London School of Economics and Political Science 1–29.
Leno, Volodymyr, Marlon Dumas, Marcello La Rosa, Fabrizio Maria Maggi, and Artem
Polyvyanyy. 2020a. “Automated Discovery of Data Transformations for Robotic Process
Automation.” arXiv preprint arXiv:2001.01007 .
Leno, Volodymyr, Marlon Dumas, Fabrizio Maria Maggi, and Marcello La Rosa. 2018. “Multi-
Perspective process model discovery for robotic process automation.” In CEUR Workshop
Proceedings, Vol. 2114, 37–45.
Leno, Volodymyr, Artem Polyvyanyy, Marlon Dumas, Marcello La Rosa, and Fabrizio Maria
Maggi. 2020b. “Robotic Process Mining: Vision and Challenges.” Business and Information
Systems Engineering .
Leno, Volodymyr, Artem Polyvyanyy, Marcello La Rosa, Marlon Dumas, and Fabrizio Maria
Maggi. 2019. “Action Logger: Enabling Process Mining for Robotic Process Automation.”
In CEUR Workshop Proceedings, Vol. 2420, 124–128.
Leopold, Henrik, Han van Der Aa, and Hajo Reijers. 2018. “Identifying candidate tasks for
robotic process automation in textual process descriptions.” In Enterprise, Business-Process
and Information Systems Modeling. BPMDS 2018, EMMSAD 2018. Lecture Notes in Busi-
ness Information Processing, edited by J. Gulden, I. Reinhartz-Berger, R. Schmidt, S. Guer-
reiro, W. Gu´edria, and P. Bera, Vol. 318, 67–81. Springer, Cham.
Leshob, Abderrahmane, Audrey Bourgouin, and Laurent Renard. 2018. “Towards a Process
Analysis Approach to Adopt Robotic Process Automation.” In 15th International Confer-
ence on e-Business Engineering, 46–53. IEEE.
Lewicki, Przemyslaw, Jacek Tochowicz, and Jeroen van Genuchten. 2019. “Are Robots Taking
Our Jobs? A RoboPlatform at a Bank.” IEEE Software 36 (3): 101–104.
Lohrmann, Matthias, and Manfred Reichert. 2016. “Effective application of process improve-
ment patterns to business processes.” Software & Systems Modeling 15 (2): 353–375.
Masood, Adnan, and Adnan Hashmi. 2019. “Cognitive Robotics Process Automation: Auto-
mate This!” In Cognitive Computing Recipes, 225–287. Apress, Berkeley, CA.
Moffitt, Kevin C., Andrea M. Rozario, and Miklos A. Vasarhelyi. 2018. “Robotic Process
Automation for Auditing.” Journal of Emerging Technologies in Accounting 15 (1): 1–10.
Mohanty, Soumendra, and Sachin Vyas. 2018. “Intelligent Process Automation = RPA + AI.”
In How to Compete in the Age of Artificial Intelligence, 125–141. Apress, Berkeley, CA.
Mundbrod, Nicolas, and Manfred Reichert. 2017. “Flexible Task Management Support for
Knowledge-Intensive Processes.” In 2017 IEEE 21st International Enterprise Distributed
Object Computing Conference (EDOC), 95–102.
Ng, Chun Kit, Chun Ho Wu, Kai Leung Yung, Wai Hung Ip, and Tommy Cheung. 2018. “A
32
semantic similarity analysis of Internet of Things.” Enterprise Information Systems 12 (7):
820–855.
Osmundsen, Karen, Jon Iden, and Bendik Bygstad. 2019. “Organizing Robotic Process Au-
tomation: Balancing Loose and Tight Coupling.” In 52nd Hawaii International Conference
on System Sciences, 6918–6926.
Patel, Maharsh, Aastha Shukla, Raunaq Porwal, and Radhika Kotecha. 2019. “Customized
Automated Email Response Bot Using Machine Learning and Robotic Process Automation.”
SSRN Electronic Journal .
Penttinen, Esko, Henje Kasslin, and Aleksandre Asatiani. 2018. “How to Choose Between
Robotic Process Automation and Back-End System Automation?” 26th European Confer-
ence on Information Systems 1–14.
Petersen, Kai, Robert Feldt, Shahid Mujtaba, and Michael Mattsson. 2008. “Systematic map-
ping studies in software engineering.” In 12th International Conference on Evaluation and
Assessment in Software Engineering (EASE) 12, 1–10.
Petersen, Kai, Sairam Vakkalanka, and Ludwik Kuzniarz. 2015. “Guidelines for conducting
systematic mapping studies in software engineering: An update.” Information and Software
Technology 64: 1–18.
Radke, Andreas M, Minh Trang Dang, and Albert Tan. 2020. “Using Robotic Process Au-
tomation (RPA) to enhance item master data maintenance process.” LogForum 16 (1):
129–140.
Reichert, Manfred, and Barbara Weber. 2012. Enabling Flexibility in Process-Aware Informa-
tion Systems: Challenges, Methods, Technologies. Springer, Berlin-Heidelberg.
Riedl, Yannik, and K. Richard Beetz. 2019. “Robotic Process Automation: Developing a Multi-
Criteria Evaluation Model for the Selection of Automatable Business Processes.” In 25th
Americas Conference on Information Systems, Cancun, .
Royce, Winston W. 1987. “Managing the development of large software systems: concepts and
techniques.” In 9th international conference on Software Engineering, 328–338.
Rutschi, Corinna, and Jens Dibbern. 2020. “Towards a framework of implementing software
robots: Transforming Human-executed Routines into Machines.” Data Base for Advances
in Information Systems 51 (1): 104–128.
Santos, Filipa, R´uben Pereira, and Jos´e Braga Vasconcelos. 2019. “Toward robotic process au-
tomation implementation: an end-to-end perspective.” Business Process Management Jour-
nal 26 (2): 405–420.
Schmitz, Manfred, Christian Dietze, and Christian Czarnecki. 2019. “Enabling digital transfor-
mation through robotic process automation at Deutsche Telekom.” In Digitalization Cases,
Management for Professionals, edited by N. Urbach and M. R¨oglinger, 15–33. Springer,
Cham.
Schmitz, Manfred, Christoph Stummer, and Gerke Michael. 2019. “Smart Automation as En-
abler of Digitalization? A Review of RPA/AI Potential and Barriers to Its Realization.”
In Future Telco, Management for Professionals, edited by P. Kr¨ussel, 349–358. Springer,
Cham.
S´eguin, Sara, and Im`ene Benkala¨ı. 2020. “Robotic Process Automation (RPA) Using an Integer
Linear Programming Formulation.” Cybernetics and Systems 51 (4): 357–369.
Steinau, Sebastian, Andrea Marrella, Kevin Andrews, Francesco Leotta, Massimo Mecella,
and Manfred Reichert. 2019. “DALEC: a framework for the systematic evaluation of data-
centric approaches to process management software.” Software & Systems Modeling 18 (4):
2679–2716.
Stople, Annette, Heidi Steinsund, Jon Iden, and Bendik Bygstad. 2017. “Lightweight IT and
the IT function: experiences from robotic process automation in a Norwegian bank.” Bibsys
Open Journal Systems 25 (1).
Suri, Vipin K, Marianne Elia, Pavan Arora, and Jos van Hillegersberg. 2018. “Automation of
Knowledge-Based Shared Services and Centers of Expertise.” In Digital Services and Plat-
forms, Considerations for Sourcing, Global Sourcing, Lecture Notes in Business Information
Processing, edited by J. Kotlarsky, I. Oshri, and L. Willcocks, Vol. 344. Springer, Cham.
33
Suri, Vipin K, Marianne Elia, and Jos van Hillegersberg. 2017. “Software bots - The next
frontier for shared services and functional excellence.” In Global Sourcing of Digital Services:
Micro and Macro Perspectives, Global Sourcing, Lecture Notes in Business Information
Processing, edited by I. Oshri, J. Kotlarsky, and L. Willcocks, Vol. 306, 81–94. Springer,
Cham.
Syed, Rehan, Suriadi Suriadi, Michael Adams, Wasana Bandara, Sander J.J. Leemans, Chun
Ouyang, Arthur H.M. ter Hofstede, Inge van de Weerd, Moe Thandar Wynn, and Hajo A.
Reijers. 2020. “Robotic Process Automation: Contemporary Themes and Challenges.” Com-
puters in Industry 115.
van der Aalst, Will M.P. 2011. Process mining: discovery, conformance and enhancement of
business processes. Springer.
van der Aalst, Will M.P., Martin Bichler, and Armin Heinzl. 2018. “Robotic Process Automa-
tion.” Business and Information Systems Engineering 60 (4): 269–272.
Wanner, Jonas, Adrian Hofmann, Marcus Fischer, Florian Imgrund, Christian Janiesch, and
Jerome Geyer-Klingeberg. 2020. “Process selection in RPA projects - Towards a quantifiable
method of decision making.” 40th International Conference on Information Systems, ICIS
2019 1–17.
Weber, Barbara, Shazia Sadiq, and Manfred Reichert. 2009. “Beyond rigidity–dynamic process
lifecycle support.” Computer Science-Research and Development 23 (2): 47–65.
Weske, Mathias. 2007. Business Process Management: Concepts, Languages, Architectures.
Springer.
Wewerka, Judith, and Manfred Reichert. 2021. “Robotic Process Automation in the Automo-
tive Industry - Lessons Learned form an Exploratory Case Study.” In 15th International
Conference on Research Challenges in Information Science, 1–17.
Willcocks, Leslie, and Mary Lacity. 2015. “Robotic Process Automation: The Next Transfor-
mation Lever for Shared Services.” The Outsourcing Unit Working Research Paper Series
15 (7): 1–35.
Willcocks, Leslie, and Mary Lacity. 2016. “Robotic Process Automation at Telef´onica O2.”
MIS Quarterly Executive 15 (1): 21–35.
Willcocks, Leslie, Mary Lacity, and Andrew Craig. 2015. “The IT Function and Robotic Process
Automation.” The Outsourcing Unit Working Research Paper Series 15 (5): 1–39.
William, Woelly, and Linda William. 2019. “Improving Corporate Secretary Productivity using
Robotic Process Automation.” Proceedings - 2019 International Conference on Technologies
and Applications of Artificial Intelligence, TAAI 2019 .
Wr´oblewska, Anna, Tomasz Stanis lawek, Bart lomiej Prus-Zaj¸aczkowski, and Lukasz Gar-
ncarek. 2018. “Robotic Process Automation of Unstructured Data with Machine Learning.”
Annals of Computer Science and Information Systems 16: 9–16.
Yatskiv, Solomiya, Iryna Voytyuk, Nataliia Yatskiv, Oksana Kushnir, Yuliia Trufanova, and
Valentyna Panasyuk. 2019. “Improved Method of Software Automation Testing Based on
the Robotic Process Automation Technology.” In 9th International Conference on Advanced
Computer Information Technologies (ACIT), 293–296. IEEE.
34
