Description: Logs recorded by RPA tools are characterized by long sequences of actions and/or events that re-flect a number of routine executions. A log can record information about several routines, whose actions and events are mixed in some order that reflects the particular order of their execution by the user (Bosco et al. 2019). In
addition, the same routine can be spread across multiple logs, making the automated identification of routines far from being trivial.
Objective: Identify the routines to be (potentially) emulated through software robots by looking at the RPA logs that keep track of the user actions taking place during a run of the system. This issue is known as “segmentation”. Approaches: One possible approach to tackle this challenge is to rely on log analysis solutions in the HumanComputer Interaction (HCI) field (Dumais et al. 2014; Dev and Liu 2017; Marrella and Catarci 2018), which focus on identifying frequent user tasks inside logs consisting of actions at different granularity. Alternately, local process mining approaches (Tax et al. 2016) or sequential pattern mining (Dong 2009) can be employed to identify sequential patterns of non-consecutive actions that tend to be repeated multiple times across multiple logs (Bosco et al. 2019). An interesting recent approach is the one of Gao et al. (Gao et al. 2019), where the authors present a learning-based approach that allows for completely automated RPA-rule deduction, on the basis of captured historical low-level user behavior. However, to date, no available solution exists that allows for automatically: (i) understanding which user actions have to be considered inside the log (separating noise from actions that contribute to routines); (ii) interpreting their semantics on the basis of their granularity and (iii) identifying which routines they belong to. Solving the above challenges would allow us to cluster all user actions associated with a routine in a well bounded execution trace. Consequently, all such execution traces would be organized into a routine-based log.
Description: While existing RPA tools allow one to automate a wide range of routines, they do not allow one to determine in an automated way which routines are good candidates for automation in the first place.
Objective: Given a list of routine-based logs, identify automatically which routines are good candidates for being automated by RPA tools.
Approaches: To date, current RPA tools provide very limited support to this challenge, which is often performed by means of interviews, walkthroughs, direct observation of workers, and analysis of documentation that may be of poor quality and difficult to understand. This manual approach allows analysts to identify the most obvious routines, while it is not suitable to detect those routines that are not executed on a daily basis or that are performed across multiple business units in different ways. The work of Jimenez-Ramirez and Reijers (Jimenez-Ramirez et al. 2019) proposes to mitigate this issue through a method to improve the early stages of the RPA lifecycle using process mining techniques (van der Aalst 2016). On the other hand, a potential concrete solution to tackle this challenge is proposed by Bosco et al. (Bosco et al. 2019), where the authors present a method to analyze routine-based logs in order to discover routines that are fully deterministic.
To this end, the method combines a technique for compressing a set of routines into an acyclic automaton, with techniques for rule mining and for discovering data transformations.
Description: In RPA tools, there is a lacking of testing environments. As a consequence, SW robots are developed through a trial-and-error approach consisting of three steps that are repeated until success (Leno et al. 2018): (i) First, a human designer produces a flowchart diagram that includes the actions to be performed by the SW robot on a target system; (ii) Second, SW robots are typically deployed in production environments, where they interact with information systems, with a high risk of errors due to inaccurate modeling of flowcharts; (iii) Third, if SW robots are not able to reproduce the behavior of the users for a specific routine, then the designer adjusts the flowchart diagrams to fix the identified gap. While this approach is effective to execute simple rule-based logic in situations where there is no room for interpretation, it becomes time-consuming and error-prone in the presence of routines that are less predictable or require some level of human judgement. Indeed, the designer should have a global vision of all possible unfoldings of the routines to define the appropriate behaviors of the SW robot, which becomes complicated when the number of unfoldings increases. In cases where the rule set does not contain a suitable response for a specific situation, robots allow for escalation to a human supervisor.
Objective: Once the routines to be automated and the user actions that constitute them (i.e., the routine-based logs) have been identified, the target is to automatically generate the flowchart diagrams describing the behaviors of the SW robots required to successfully execute the routines.
Approaches: A possible solution to this challenge can be to resort on discovery algorithms from the process mining field (van der Aalst 2016) and to automatically extract flowcharts in the form of Petri nets/BPMN models from routine-based logs. Thus, it is necessary to investigate from the literature on process discovery (Augusto et al. 2019) which algorithms suit better to extract the base structure of flowchart diagrams from a routine-based log. Since such discovered flowcharts will reflect real routine executions, they may contain/miss (un-)necessary user actions with respect to the SW robot’s expected behavior. To mitigate this issue, it is possible to leverage another process mining technique, named trace alignment (Adri- ansyah, Sidorova, and van Dongen 2011), which would allow us to properly clean the discovered flowchart diagrams, by filtering out the unwanted behaviors found in the previous discovery stage.
Description: In modern contexts, human operators usually enact not just single tasks but complex workflows, consisting of many interrelated routines. However, the current RPA technology allows for developing SW robots for executing single, independent routines. Only manual support is provided to orchestrate multiple routines, i.e., the
Figure 1: Overview of the pipeline of potential approaches required to tackle the research challenges
management of more complex workflows is completely delegated to human supervisors.
Objective: Automated generation of RPA workflows consisting of many interrelated routines.
Approaches: To synthesize complex workflows through an intelligent orchestration of the robots’ routines, automated planning techniques in AI can be employed (Ghal- lab, Nau, and Traverso 2004). The application of planning for tackling the composition issue has been already proved to be successful in real world domains (Marrella 2019). The idea is to consider the robots’ routines as black boxes, i.e., as planning actions with specific preconditions and effects, and to delegate to a planning system the generation of a proper strategy to automatically compose them in a larger workflow that coordinates their orchestration.
Figure 1 shows a graphical overview of the pipeline of potential approaches required to tackle the four identified research challenges to achieve the aforementioned objectives.
RPA recently gained a lot of attention in the BPM domain (van der Aalst, Bichler, and Heinzl 2018). Since RPA operates at the UI level, rather than at the system level, it allows one to apply automation without any changes in the underlying information system. Thus, the entry barrier of adopting RPA in BPs that are already in place is lower compared to conventional BPM (Gao et al. 2019). However, the current generation of RPA tools is driven by predefined rules and manual configurations made by expert users rather than by AI (Lohr 2018), preventing a widespread adoption of these tools in the BPM domain. In this paper, we have tackled this issue starting from an in-depth experimentation of the RPA tools available in the market. Then we have provided a classification framework to categorize them on the basis of some key dimensions and we have derived four research challenges and discussed potential approaches necessary to inject intelligence into the current RPA technology, from a BPM perspective. It is worth to notice that, according to Table 1, the logs produced by the tested RPA tools have a poor quality (ac-
tions may be missing or not recorded properly), since they are mainly used for debugging purposes. Increasing the quality of RPA logs is a fundamental prerequisite to properly tackle the proposed research, which leverages a log analysis to discover, identify, model and compose routines in an automated way. To this end, RPA tools should aim at logs at the highest possible quality level.
To mitigate this issue, we are currently developing an action logger to be attached to the existing RPA tools, in order to enable the creation of routines-based logs of an acceptable quality. Apart from the need to increase the quality of RPA logs, as a future work, this research aims at also improving the auditability (RPA logs are auditable), upgradability (flowchart diagrams describing SW robots’ behavior will be always updated to the current state of the system execution) and the resiliency (SW robots will be always upgraded to deal with new behaviors, making them very robust to any contextual change that may arise during a routine execution) of SW robots. Furthermore, scalability must be improved as well. Human capacity is difficult to scale in situations where demand fluctuates, instead SW robots operate at whatever speed is demanded by the work volume.
To conclude, we note that our study has a threat to validity, since we analyzed only a sample of the RPA tools available on the market. As a consequence, our findings can not be generalized beyond the scope of the tested RPA tools. Nonetheless, we consider this work as an important first step towards the realization of intelligent solutions for RPA. Moreover, we also envision that this research will provide long-term benefits on the companies workforce, e.g., by improving the customer service in the front office while at the same time reducing the back office tasks.
[Adriansyah, Sidorova, and van Dongen 2011] Adriansyah, A.; Sidorova, N.; and van Dongen, B. F. 2011. Cost-Based Fitness in Conformance Checking. 2011 Eleventh International Conference on Application of Concurrency to System Design 57–66.
[Aguirre and Rodriguez 2017] Aguirre, S., and Rodriguez, A. 2017. Automation of a Business Process Using Robotic
Process Automation (RPA): A Case Study. In Applied Com-
puter Sciences in Engineering, 65–71. Cham: Springer In-
ternational Publishing.
[AI-Multiple 2019] AI-Multiple. 2019. All 52 RPA Soft- ware Tools and Vendors: Sortable List [2019]. https://blog. aimultiple.com/rpa-tools/.
[Augusto et al. 2019] Augusto, A.; Conforti, R.; Dumas, M.; Rosa, M. L.; Maggi, F. M.; Marrella, A.; Mecella, M.; and Soo, A. 2019. Automated Discovery of Process Models from Event Logs: Review and Benchmark. TKDE 31(4).
[Bisbal et al. 1999] Bisbal, J.; Lawless, D.; Wu, B.; and Grimson, J. 1999. Legacy information systems: Issues and directions. IEEE Software 16(5):103–111.
[Bosco et al. 2019] Bosco, A.; Augusto, A.; Dumas, M.; La Rosa, M.; and Fortino, G. 2019. Discovering Automatable Routines From User Interaction Logs. In 17th International Conference on Business Process Management (BPM’19), Forum track, Vienna, Austria, 144–162. Cham: Springer International Publishing.
[Dev and Liu 2017] Dev, H., and Liu, Z. 2017. Identifying Frequent User Tasks from Application Logs. In Proceedings of the 22Nd International Conference on Intelligent User Interfaces, IUI ’17, 263–273. New York, NY, USA: ACM.
[Dong 2009] Dong, G. 2009. Sequence Data Mining. Berlin, Heidelberg: Springer-Verlag.
[Dumais et al. 2014] Dumais, S.; Jeffries, R.; Russell, D. M.; Tang, D.; and Teevan, J. 2014. Understanding User Behavior Through Log Data and Analysis. In Ways of Knowing in HCI. New York, NY: Springer. 349–372.
[Gao et al. 2019] Gao, J.; van Zelst, S. J.; Lu, X.; and van der Aalst, W. M. P. 2019. Automated robotic process automation: A self-learning approach. In On the Move to Meaningful Internet Systems: OTM 2019 Conferences, 95–112. Cham: Springer International Publishing.
[Geyer-Klingeberg et al. 2018] Geyer-Klingeberg, J.; Nakladal, J.; Baldauf, F.; Veit, F.; van der Aalst, W.; Casati, F.; Conforti, R.; de Leoni, M.; and Dumas, M. 2018. Process Mining and Robotic Process Automation: A Perfect Match. In 16th International Conference on Business Process Management (BPM’18), Dissertation/Demos/Industry track, Sidney, Australia, 124–131.
[Ghallab, Nau, and Traverso 2004] Ghallab, M.; Nau, D.; and Traverso, P. 2004. Automated Planning: theory and practice. Elsevier.
[Hill, Ford, and Farreras 2015] Hill, J.; Ford, W. R.; and Far- reras, I. G. 2015. Real conversations with artificial intelligence: A comparison between human–human online conversations and human–chatbot conversations. Computers in Human Behavior 49:245–250.
[Jimenez-Ramirez et al. 2019] Jimenez-Ramirez, A.; Reijers, H. A.; Barba, I.; and Del Valle, C. 2019. A Method to Improve the Early Stages of the Robotic Process Automation Lifecycle. In 31st International Conference on Advanced Information Systems Engineering (CAiSE’19), Rome, Italy, 446–461. Cham: Springer International Publishing.
[Kirchmer 2017] Kirchmer, M. 2017. Robotic Process Automation-Pragmatic Solution or Dangerous Illusion. BTOES Insights, June’17.
[Lacity, Willcocks, and Craig 2015] Lacity, M.; Willcocks, L. P.; and Craig, A. 2015. RPA at Telefonica O2. The London School of Economics and Political Science.
[Leno et al. 2018] Leno, V.; Dumas, M.; Maggi, F. M.; and La Rosa, M. 2018. Multi-perspective process model discovery for robotic process automation. CAiSE Doct. Cons.
[Leno et al. 2019] Leno, V.; Polyvyanyy, A.; Rosa, M. L.; Dumas, M.; and Maggi, F. M. 2019. Action logger: Enabling process mining for robotic process automation. In Proceedings of the Dissertation Award, Doctoral Consortium, and Demonstration Track at 17th International Conference on Business Process Management, (BPM’19), Vienna, Austria, 124–128.
[Lohr 2018] Lohr, S. 2018. The Beginning of a Wave: A.I. Tiptoes Into the Workplace. https://www.nytimes.com/ 2018/08/05/technology/workplace-ai.html/.
[Marrella and Catarci 2018] Marrella, A., and Catarci, T. 2018. Measuring the Learnability of Interactive Systems Using a Petri Net Based Approach. In Proceedings of the 2018 Designing Interactive Systems Conference, DIS ’18, 1309–1319. New York, NY, USA: ACM.
[Marrella 2019] Marrella, A. 2019. Automated Planning for Business Process Management. J. Data Semantics 8(2):79– 98.
[Reichert and Weber 2012] Reichert, M., and Weber, B. 2012. Enabling Flexibility in Process-Aware Information Systems - Challenges, Methods, Technologies. Springer Berlin Heidelberg.
[SE 2019] SE, C. 2019. Academic Alliance. https://www. celonis.com/academic-alliance.
[Tax et al. 2016] Tax, N.; Sidorova, N.; Haakma, R.; and van der Aalst, W. M. 2016. Mining local process models. Journal of Innovation in Digital Ecosystems 3(2):183–196.
[Tornbohm 2017] Tornbohm, C. 2017. Gartner market guide for Robotic Process Automation software. Report G00319864. Gartner.
[van der Aalst and van Hee 2004] van der Aalst, W., and van Hee, K. 2004. Workflow management: models, methods, and systems. MIT press.
[van der Aalst et al. 2012] van der Aalst, W.; Adriansyah, A.; de Medeiros, A. K. A.; Arcieri, F.; Baier, T.; Blickle, T.; Bose, J. C.; van den Brand, P.; Brandtjen, R.; Buijs, J.; Burattin, A.; Carmona, J.; Castellanos, M.; Claes, J.; Cook, J.; Costantini, N.; Curbera, F.; Damiani, E.; de Leoni, M.; Delias, P.; van Dongen, B. F.; Dumas, M.; Dustdar, S.; Fahland, D.; Ferreira, D. R.; Gaaloul, W.; van Geffen, F.; Goel, S.; G¨unther, C.; Guzzo, A.; Harmon, P.; ter Hofstede, A.; Hoogland, J.; Ingvaldsen, J. E.; Kato, K.; Kuhn, R.; Kumar, A.; La Rosa, M.; Maggi, F.; Malerba, D.; Mans, R. S.; Manuel, A.; McCreesh, M.; Mello, P.; Mendling, J.; Montali, M.; Motahari-Nezhad, H. R.; zur Muehlen, M.; MunozGama, J.; Pontieri, L.; Ribeiro, J.; Rozinat, A.; Seguel P´erez, H.; Seguel P´erez, R.; Sep´ulveda, M.; Sinur, J.; Soffer, P.;
Song, M.; Sperduti, A.; Stilo, G.; Stoel, C.; Swenson, K.; Ta-
lamo, M.; Tan, W.; Turner, C.; Vanthienen, J.; Varvaressos,
G.; Verbeek, E.; Verdonk, M.; Vigo, R.; Wang, J.; Weber, B.;
Weidlich, M.; Weijters, T.; Wen, L.; Westergaard, M.; and
Wynn, M. 2012. Process Mining Manifesto. In Business
Process Management Workshops, 169–194. Berlin, Heidel-
berg: Springer Berlin Heidelberg.
[van der Aalst, Bichler, and Heinzl 2018] van der Aalst, W. M. P.; Bichler, M.; and Heinzl, A. 2018. Robotic Process Automation. Business & Information Systems Engineering 60(4):269–272.
[van der Aalst 2016] van der Aalst, W. M. P. 2016. Process Mining: Data Science in Action. Heidelberg: Springer, 2 edition.
[Willcocks, Lacity, and Craig 2015] Willcocks, L. P.; Lacity, M.; and Craig, A. 2015. The IT Function and Robotic Process Automation. The London School of Economics and Political Science.
[Willcocks 2016] Willcocks, L. 2016. Service Automation : robots and the future of work. Warwickshire, United Kingdom: Steve Brookes Publishing.