1 Introduction

Business processes are often supported by Business Process Management Systems (BPMSs), a type of Information System that manages and coordinates the execution of the business process and the allocation of tasks to resources in the process. Based on a design time specification of the tasks and their order, and the organizational structure, the BPMS selects at run time the optimal resource for the execution of a task for a certain process instance. This allocation mechanism is often termed (dynamic) role resolution, as the allocation of a resource is based on the role and position the resource has in the organization (e.g. clerk of the financial department).

Current resource models and role resolution mechanisms are rather basic and limited in their specification [1]. They focus on single human actors per activity, only take into account basic organizational features of a resource (e.g. role, position, department), and do not consider case information in the allocation of tasks to resources. In this position paper we propose ideas to extend and advance the resource modeling and role resolution capabilities of BPM technology specifically to enhance human and cognitive aspects of the system. We propose a.o. to extend the resource model with a resource’s characteristics such as its capabilities, experience and expertise and to use this information in role resolution in order to improve e.g. the quality of work, satisfaction, motivation, learning and training of employees.

The structure of this position paper is as follows. First, the current state-of-the-art in role resolution is discussed in Sect. 2. Next, extensions towards human and cognitive aspects are discussed and examples illustrating the opportunities and advanced requirements for role resolution are presented. Finally, Sect. 4 presents a brief research agenda and Sect. 5 concludes this paper.

2 Dynamic Role Resolution

BPMSs support and automate business processes by leading process instances (also called cases) through the tasks of the process in the right order and by coordinating the resources that execute these tasks. A resource is an entity that is assigned to a task and is requested at runtime to perform a work item for a specific process instance in order to complete the objective of the task for that case [1].

At the basis for this type of process support are two models: the process model specifies which tasks have to be executed in which order, and the resource model specifies the resources that are involved in the execution of tasks in the process, their mutual relationships (who can replace who) and the organizational structure. These models are specified at design time. In the process model it is indicated per task which type of resource is allowed to execute the task, linking to the organizational structure defined in the resource model. For instance, in Fig. 1 task B has to be executed by an employee of role R1 and department G2. The specific resources that may execute task A are employees Clare and Jody. At run time, when specific process instances are executed, one of these specific resources is allocated to the work item.

Fig. 1.
figure 1

A process model and accompanying resource model (taken from [2])

Mostly, the resource models used in contemporary BPM technology and theory are rather simple: resources belong to an organizational structure (group or department) and execute a certain role (e.g. clerk, expert, manager). Role resolution mechanisms are limited to this 2-dimensional organizational view. This paper proposes a number of (conceptual) extensions to this limited view in order to improve the match between the case and the resource and to enhance human aspects of work allocation.

Some researchers, however, have focused on further describing the allocation of work items to resources from a technical perspective. [1] elaborates on assignment and synchronization policies that determine in which way the work items from a case are distributed and assigned to employees. They distinguish for instance between push (the system forces the resource to start working on a certain item) and pull (the resource requests the next work item from the system) patterns, individual or group work item inboxes, whether or not delegation to another resource after assignment is possible, and whether the queuing of work items is done in a queue (fixed order) or pool (free selection). Also, [3] presents a number of resource patterns that describe the technical features in BPMSs mainly driven by the lifecycle of a work item. They evaluate which of these theoretical patterns are actually supported by commercially available BPMSs. And they also discuss a number of standard allocation rules such as round robin, shortest work queue, selection of the same resource as for the previous activity for the case, and the ‘four eyes principle’. Finally, some more recent work focuses on the optimization of work item/resource allocation with respect to process performance indicators such as waiting time, throughput time and resource utilization, e.g. [4], and on most optimal work-handovers [5]. Neither of these prior works, however, focus on a more advanced specification of work item allocation to resources with respect to cognitive and human aspects.

3 Proposed Advancements

In order to advance role resolution mechanisms, we propose extensions to the current state-of-the-art, being inspired by a number of practical BPM research projects [7, 8]. Figure 2 graphically depicts these extensions that are discussed one by one here.

Fig. 2.
figure 2

Overview of ingredients for advanced role resolution.

Resource Characteristics - The current practice often considers a resource to be an individual human being with a certain organizational position that allows him or her to execute certain tasks in a process. In practice, especially in application domains outside the administrative domain such as manufacturing and healthcare, also different types of resources exist: teams consisting of several experts (e.g. surgery teams with a surgeon, anesthesiologist, and nurses), non-human resources (e.g. robots, web services), etc. In order to be suitable candidates for the execution of a task, these resources have to satisfy a number of criteria (task requirements) that are not included in the organizational model. The different resource types may be interchangeable (e.g. a robot and a human may both execute the same task). Furthermore, human resources are currently selected based on their organizational position only. However, humans in the same position are never exactly the same. They have e.g. different backgrounds, capabilities, experience, preferences and personal goals. These additional characteristics may be taken into account when allocating a resource: for instance, a clerk with many years of experience executing a task may be more suitable to handle a difficult case than a clerk who has just started. Or the allocation of work items may be matched with personal learning goals of the employees.

Case Characteristics - In none of the existing allocation approaches the characteristics of the cases that run through the process are taken into account. Of course, these cases all have something in common – they have to be handled by the same process – but still many specific characteristics such as history, personal details of the client, and details of the case (e.g. the amount of a loan, credibility category of the applicant, etc.) can be taken into account to match the case with the best resource to execute a task. This may also improve human and cognitive aspects of the system. For instance, a request for a high amount of loan may be assigned to an expert employee to avoid mistakes (good for both satisfaction and external quality), or a case may be assigned to the employee that has communicated with the client before (in the context of another case or process) leading to higher familiarity and less set-up time.

Process Objectives - The current role resolution mechanisms, including the proposed extensions above, focus on the operational allocation of resources. On top of that, one may also consider the goals of the process on the tactical level to improve the overall process. Apart from the already researched process performance (e.g. shortest waiting time) [5, 6], there may also be other objectives for the overall management of the process. For instance, by using the detailed case characteristics and resource capabilities, one may be able to match cases based on training purposes for the human resources (once in a while the resource gets a more difficult case to execute), or improve work variety for the human resource (the activities to be executed and the cases to be executed differ enough such that the resource doesn’t feel bored). In addition to that, the set of general allocation rules, such as the ‘four-eyes principle’, may be extended with rules based on historical information and case and resource specific information (e.g. a task that is disapproved always should be sent back to the resource that originally handled it such that this resource can learn from mistakes).

These proposed extensions to the current state-of-the-art in role resolution are illustrated with a few examples from practice that show the possibilities of improving cognitive and human aspects in the process by advancing work item allocation.

Example 1 (inspired by [6]) – The invoicing department of a telecommunications company has to process many invoices received from national and international suppliers. These invoices can be of several types (from a preferred supplier or not, based on a purchase order or not, etc.). All invoices are handled by the same process and are therefore considered the same, but after an analysis of the process performance, it turns out that many mistakes are made. The reason for this is that not all employees are well-trained and experienced for all types of invoices. The inexperienced workers for instance make many mistakes with the VAT of international invoices. When the characteristics of the invoices and the capabilities of the resources are better mapped, the matching between the resource and the invoice could be improved. First of all, one could think of only assigning invoices to resources that are known to handle these correctly, but in a later stage one could also try to improve the work variety of the employees by offering alternating types of invoices and prevent workers from boredom, or train and help the resources build new skills by assigning more difficult invoices once in a while.

Example 2 (inspired by [7]) – a financial company offers personal loans to clients via the internet. In order to request a loan the (prospected) client has to fill an online form and add details about himself, his financial situation and the requested loan. When this online form is received, some automatic checks are performed and a call agent calls back the requestor to discuss the request, complete the information and to explain the procedure that will follow. This call agent appears to have a large impact on the success rate of eventually getting the loan offer accepted. Therefore, a good match between the request/requestor and the call agent is essential. For instance, historical data showed that a male applicant best could be called by a female agent, that the client should preferably be addressed in his native language, that call agents with a lot of experience were more successful with borderline cases, and that the time of calling back related to the time of the online request also made a difference (the business rule of calling back within an hour from the request was abandoned).

4 Research Challenges

In the previous section, we have explained our proposed conceptual advancements of the current practice in role resolution in order to create better matches between cases and resources, focusing on human aspects in the allocation. This section briefly discusses the research challenges to implement the above ideas in BPM technology.

At design time, the process, organizational model, but also the case characteristics, resource capabilities, experience, goals, etc., and the allocation rules and process objectives have to be modeled. Therefore, current process modeling languages have to be extended in order to include this information. In some cases this may be a simple information element added to an existing modeling language, but especially for modeling case characteristics and resource characteristics new models have to be created. We may build on recent developments in the area of resource modeling [8] experience and skill modeling [9, 10] and plan to investigate insights from the field of work- and organizational psychology.

Secondly, at run time, the process execution engine also has to interpret the additional information that was stored in the extended models (i.e., resource and case characteristics, allocation rules, and performance) and select the optimal resource. This requires a more advanced decision algorithm that can deal with the operational as well as the tactical process objectives. Also, the process execution engine has to deal with possible exceptions, such as sudden unavailability of a resource, and re-assign to another resource on the fly in a more advanced way.

5 Conclusion

In this paper, we have proposed a number of extensions to the current state-of-the-art in role resolution in BPMSs. The main goal of these proposals is to make a better match between the case and the resources executing tasks for the case and by that improve the process performance and process outcome (external quality), but also human and cognitive aspects of the system such as work motivation, satisfaction (internal work quality) and training. The ideas presented are mainly conceptual and still have to be elaborated upon and implemented in a (prototype) BPMS.