Workflow Management Systems and Agents - Do they Fit
Together?
Pavlos Delias
Department of Production Engineering and Management
Technical University of Crete, Univeristy campus D5 building, Chania, Greece
Abstract. Workflow management systems are an emerging category of
information systems, currently under dynamic evolution. On the other hand
software agents is a distinct research area and an also emerging paradigm for
information systems design and development. In this paper, I outline the major
points of a doctoral thesis that will focus on the intersection of these two fields.
I try to clarify the thesis specific objectives and describe the motivation
underneath. The general methodology as well as some initial findings are also
described.
1 Introduction
Workflow Management Systems emerged in the Information Systems landscape as a
promising office information systems technology at the 70s. During the 80s, they
have evolved into enactment machines of operational models. Their critical feature of
that time was that they were too rigid to support the integration of human activities.
This essential requirement advantaged the development of systems that could support
collaborative work. Singh and Huhns [1] support that “Workflows have been with us
from the dawn of time” and sectionalize the systems into five generations: Starting
from the “manual” ones which were a side-effect of bureaucracy; they continue with
the “closed” ones that focused mainly on data processing and on the automation of the
existing manual activities. The third generation concerned the “database-centric”
systems. It was then when data and process appeared to decouple themselves. The
next generation refers to the current situation. This generation’s systems provide the
separation of control from the application. Finally, Singh and Hunhs predict that the
next generation will incorporate agent-based systems.
Abott and Sarin [2] provided a different taxonomy of the WFMS. They name as “first
generation” systems the systems that were “application-specific”. Those systems were
tightly related to specific functions (e.g. document management) and they were closed
and proprietary. During the second generation, the workflow logic is separated from
the application one; while the integration of third-party tools becomes available.
Current situation is mapped on the third generation. Contemporary WFMS provide
access to other applications through APIs and they integrate third-party tools as well.
They adopt standards-based architectures and they become far more user-friendly.
Abott and Sarin’s prediction for the next generation describes a ubiquitous
environment; interchange of data and control is the focal event. Sheth and his
Proceedings of ICEIS 2008
Tenth International Conference on Enterprise Information Systems
Copyright © INSTICC
colleagues [3] illustrated the evolution of the WF runtime system architectures.
Starting from centralized / one-engine early systems, the architectures evolved to
more distributed ones; including web-orientation and mobile-agents enhancements.
As depicted in [3] the evolution will continue by supporting organic processes. In [4]
a very explanatory figure demonstrating the history of automation and workflow
systems is provided.
My thesis is trying to keep a pace with the technological advancement of WFMS and
strives to make a contribution towards more open and ubiquitous workflow
architectures. The specific objectives and the research problem I am focusing on are
described in the next section, while in section three I briefly discuss existing work,
research trends and motivation. In the last section, I present the work made so far and
discuss the next steps and how I am planning to achieve them.
2 Objectives of the Research
2.1 General Objectives
Different types of theses have been identified in the literature (Table 1). My thesis is
dedicated in providing a unifying framework and providing a new tool for verification
purposes.
Table 1. Types of Ph. D. Theses.
Opens up new area Produces an
ambitious system
Provides unifying
framework
Provides empirical
data
Resolves long-
standing question
Derives superior
algorithms
Thoroughly explores
an area
Develops new
methodology
Contradicts existing
knowledge
Develops a new tool
Experimentally
validates theory
Produces a negative
result
More specifically, my thesis is situated in the intersection of two fields: Workflow
Management Systems and Software Agents. It tries to unify these fields by examining
their interactions. Although both software agents and workflow management are
established areas of research, few works focus on their intersection. The overall
objective of the thesis is to identify the agent contribution potentials in WFMS and to
provide efficient solutions in those niches.
2.2 Research Problem
Following the general objectives defined in the previous paragraph, agent contribution
potentials to WFMS are identified and classified. For the classification criteria we
adopt the concept and the standards of WfMC [5]. The reference model that WfMC
provides is broadly accepted in an area where a bold confusion of standards exists.
Adopting a popular reference model will hopefully make the thesis contribution more
identifiable and easier disseminated.
The first challenge is to propose a functional decomposition along the reference
model of WfMC. Then, for each function, existing approaches should be categorized.
Every function is a distinct operational utility of WFMS (e.g. scheduling, task
assignment, resource allocation, etc.), thus established methodologies from
operational research could be exploited to provide efficient solutions. The major
endeavor is then to adjust OR methodologies into an agent-oriented workflow
management system architecture.
3 Research Agenda – State of the Art
3.1 Trends and Standards
The term “workflow” is overloaded to the point where it is hard to distinguish what a
WFMS is meant to achieve. This happens mainly, because there is a variety of
scenarios where workflow technology is applied: diverging from Human WF to
document management; Business Rule-Driven WF; ISO certification claim; process
controlling; composite WF for Service Oriented Architectures; groupware; grid
computing; enterprise application integration, just to name a few.
Due to its interdisciplinary nature, workflow research cuts a generous swath across
many fields. Storh et al. [6] classify the active research efforts into 3 categories:
Technical issues; Management and organizational issues; and market, economic and
social issues. Li et al. [7] discern two trends in current workflow research community.
One trend embraces the Web services paradigm and strives to develop WS-related
architectures and methodologies (choreography, orchestration, Process definition
exchange, service discovery, message exchange, coordination). The other focuses on
overcoming the limitations of traditional workflow management concerning
adaptability and flexibility.
3.2 Requirements and Limitations of Existing WFMS
WFMS are currently an active field of enterprise information systems, thus some
functional requirements that could put added value are identifiable: WFMS should
find a way to manage the dynamic nature of business processes. As business
processes become more volatile; and as they start crossing the organization’s
boundaries, their interactions need a rather sophisticated supervisor. Within business
processes, many tasks are interrelated; responsibilities and data are distributed [8, 9].
This natural concurrency demands efficient techniques for task assignment; resource
allocation and scheduling. Moreover, in the case of multiple service providers, the
WFMS should be able to semantically discover the appropriate service providers;
negotiate with them and finally allocate them the work. Failures and exceptions must
be tackled adaptively and efficiently.
Contemporary WFMS must be able to operate in a pervasive computing environment.
They should be able to integrate external applications; other WFMS; heterogeneous
devices and legacy systems. Operating in the web appears a sine qua non
requirement; while supporting the users with friendly and customizable interface
would promote their application. Scalability; security and reliability still remain
critical requirements.
Considering the above requirements, many researchers have exposed the limitations
of existing systems [9-17]. WFMS lack of adaptability: most of them require an a
priori representation of a business process and all potential deviations from that
process [13]. They can not response in a reactive way to exceptions that may occur
during the execution of a process instance. They are unable to cope with dynamic
changes in resource levels and task availability, as they tend to lack the necessary
facilities to redistribute work items automatically as and when required [11]. They
suffer from disadvantages such as not supporting the dynamic
incorporation/modification of process models; poor adaptability of process models at
runtime and they are incapable of integrating distributed process models [18]. The
static workflow definition and its passive interpretation do not allow WFMS to
demonstrate flexible behavior and to deal with real-life situation such as fast changing
customer requirements and enterprise goal shifts [15, 19].
WFMS lack of resources management facilities [11, 13, 16]. They focus on the
administration[20] of processes and they pay less or even hardly any attention to the
problems such as the resource allocation and the resource restriction [20]. Resources
conflict is seldom monitored as WFMS tend to manage independently resources in an
organization. This kind of conflicts may become even more critical in the case of
cross-organizational workflows. In addition, tasks are associated with users (actors)
rather than roles [10]. Role management is a feature that still does not exist in many
systems.
Authors of [13, 21-24] noticed very early that semantics is a feature that can lift up
workflow functionality and that existing systems lack of them. Through the use of
semantics the decisions will be further automated; negotiation among actors will be
enabled; optimization of processes and learning features will be disposable; and
compensation activities will have a formal basis to lie on. Unfortunately, the use of
semantics is still in infantile level of integration in existing WFMS. They have a weak
support of correctness and reliability [25]; inadequate exception handling [11, 12];and
limited or non-existing optimization features.
Existing WFMSs tend to be centralized while their runtime systems are based on the
client-server model [26]. Relying on one central control does not allow systems to
support reliable and consistent process execution with acceptable failure resiliency,
performance, and scalability. WFMSs operate in splendid isolation and they represent
islands of automation that provide inflexible tactical solutions [14]. They lack of
heterogeneity [13] and they have poor support of interoperability [25]. Although
WfMC strives to establish generic interfaces and to enable interoperability, when
WFMSs need to exchange data they use distinct APIs calls [16]. This fact limits
significantly their extensibility [9].
3.3 Why Use Agents?
Agents are not the panacea for all the WFMS problems and limitations. Yet, they
constitute an attractive metaphor that advances WF development.
In [27], Lange and Oshima promote the use of mobile agents in the distributed
systems field by demonstrating seven arguments. In the same paper, they present a
few application areas where the agents’ paradigm could flourish (workflow is indeed
included). Mobility infuses agents with the ability of migration. This potential allows
one to decentralize a WFMS [28] and exploit the benefits of both distributed WFMS
[25, 29, 30] and of the agents paradigm in distributed systems [27].
By their nature, agents support heterogeneity. By using an abstract communication
and coordination level, they can be incorporated into the varying hardware and
operating systems architectures that dwell in a business process [28]. This enhanced
coordination ability allow agents to act as configuration facilitators [31, 32] and
advances them as a promising technology for application integration [33].
Agents modular nature can provide highly reusable workflow architectures [34]
which not only are an alternative technology to existing workflow systems but most
importantly, they also offer an alternative vision of how organizations can be
structured and managed [13].
Agents (being autonomous) can relief WF engines from some computation.
Consequently the engines’ workloads shall be reduced favoring significantly WFMS
scalability [35]. They enable the recovery process as they are stateful entities,
contributing significantly to the fault tolerance of the system. The encapsulation of
state also supports the asynchronous execution of a business process, a popular case
when human participants are involved [28]. As a more general contribution, we may
notice that the agent paradigm supports the vision of human substitution: the inherent
autonomy of software agents can fulfil activities on behalf of human with an expected
quality of service.
Another core feature of agents, reactivity, provide them with an intrinsic capability to
adapt to dynamic changes in the environment [34]. Actions do not need to be rigidly
prescribed as agents can anticipate their environment and timely as well as efficiently
respond to the changes that occur [9, 36]. Besides reactivity, pro-activeness can boost
agents’ intelligence. Agents can adopt feedback mechanisms to guide themselves
during future actions [9]. They can implement intelligent decision-making techniques
such as negotiation [8]; semantics [16, 37, 38]; planning [18, 39]. Moreover, agents
would be able to perform optimization tasks as routing and scheduling [35, 40]; task
assignment [41]; resource allocation [10]. In [20], Qiu et al. advocate that problems
such as resource collision and low efficiency of resource utilization can not be
readily addressed unless agents join the system.
Of course, designing an agent-based system is far more complicated than relying on a
traditional WFMS. One shall always balance the trade-off between design and
development complexity and efficiency and effectiveness. Let us provide a list of
cases when the agent paradigm appears to be an eminently suitable technology for
workflow management:
Process definitions can not describe entirely the problem solution [8] or can not
predict all possible paths of the process execution.
Interactions among tasks and/or participants are fairly sophisticated [8]
Applications that are modular; decentralized; and changeable [42]
The environment demands asynchronous communication [43]
The environment is radically heterogeneous
The applications call for extensive human participants integration [28]
4 Expected Outcome
As mentioned in section 2.2 the initial phase of the thesis is to identify the functions
of a WFMS where agents could contribute. These functions are categorized according
to the WfMC’s reference model and presented in Table 2. Although the proposed
classification schema needs justification, it would be out of the scope of this paper,
hence omitted.
In the ideal situation the ultimate product will be an agent-oriented workflow
management system that will demonstrate how agents can add value to every function
mentioned. In a more realistic scenario, it is possible to design an operable agent-
oriented architecture that would exploit the advantages of agents and overcome some
of the limitations mention in section 0.
In [44], we present a first proof of this vision: an agent-based workflow architecture
exploits an efficient algorithm for dispatching tasks. A robust mathematical reasoning
model is employed and it allows agents to optimally distribute the workload. We have
reasons to believe that this reasoning model could be expanded in order to address
more workflow basic functions (resource allocation; scheduling). We plan to base the
system’s development on this reasoning model so that the system will be as much
autonomous as possible. Hopefully, a consistent reasoning model will guide both the
administration of the business processes and agents coordination. In addition, we plan
to “agentify” heterogeneous devices (e.g. PDAs) so that they could be integrated into
the unifying agent architecture of the system. Finally, business process modelling
standards will be adopted, in order to facilitate the system’s interoperability.
Concluding this paper, agent paradigm seems to have a large potential in the
workflow area. Still, their integration into WFMS is not straightforward. In fact, if
there has to be a trade-off between workflow functionality and exploiting the agents’
technology, then probably the latter argument will not be favoured. So, this thesis
should provide a unifying framework that will demonstrate that agents not only they
do not limit workflow functionality, but indeed, they enhance workflow operations.
Acknowledgements
This work is part of the 03ED375 research project, implemented within the
framework of the “Reinforcement Programme of Human Research Manpower”
(PENED) and co-financed by National and Community Funds (75% from E.U.-
European Social Fund and 25% from the Greek Ministry of Development-General
Secretariat of Research and Technology).
Table 2. Functions of WFMS where agents could contribute.
Interface 1: Process
Definition Tools
Analyze; model; compose; describe; and document a BP
Process Definition Write / Edit
Definition Retrieval
Interface 2: Workflow
Client Applications
Worklist Handling
Process Control
Data Handling
User Interface
Interface 3: Invoked
Applications
Worklist Handling
Process Control
Data Handling
Service Discovery
Interface 4: Workflow
Interoperability
Common Interpretation of Process Definition
Control Information Interchange
Data Interchange
Interface 5:
Administration &
Monitoring Tools
User / Role Management
Audit Management
Resource control
Process Monitoring
Workflow Enactment
Service
Runtime Control Environment
Definition Interpretation
Execution of Tasks
Scheduling
Data Functions
Task Assignment
Resource Allocation
References
1. Singh, M.P. and M.N. Huhns, Multiagent systems for workflow. International Journal of
Intelligent Systems in Accounting, Finance & Management, 1999. 8(2): p. 105-117.
2. Abbott, K.R. and S.K. Sarin. Experiences with workflow management: issues for the next
generation. in Proceedings of the 1994 ACM conference on Computer supported
cooperative work. 1994. Chapel Hill, North Carolina, United States: ACM Press.
3. Sheth, A.P., W. Van Der Aalst, and I.B. Arpinar, Processes driving the networked
economy. IEEE Concurrency, 1999. 7(3): p. 18-31.
4. Muehlen, M.z., Workflow-based Process Controlling, ed. J. Becker. 2004, Berlin: Logos
verlag. 282.
5. WFMC, WfMC Standards, Workflow Reference Model D. Hollingsworth, Editor. 1995,
Workflow Management Coalition.
6. Stohr, E.A. and J.L. Zhao, Workflow Automation: Overview and Research Issues.
Information Systems Frontiers, 2001. 3(3): p. 281–296.
7. Li, H. and Z. Lu. Decentralized workflow modeling and execution in service-oriented
computing environment. in Service-Oriented System Engineering, 2005. SOSE 2005. IEEE
International Workshop. 2005. Beijing, CHINA.
8. Jennings, N.R., T.J. Norman, and P. Faratin, ADEPT: an agent-based approach to business
process management. ACM SIGMOD Record, 1998. 27(4): p. 32-39.
9. Jennings, N.R., et al., Autonomous Agents For Business Process Management. Applied
Artificial Intelligence, 2000. 14(2): p. 145-189.
10. Fakas, G. and B. Karakostas, A workflow management system based on intelligent
collaborative objects. Information and Software Technology, 1999. 41(13): p. 907-915.
11. Judge, D.W., et al., Agent-enhanced Workflow. BT Technology Journal, 1998. 16(3): p. 79-
85.
12. Muller, R., U. Greiner, and E. Rahm, AGENTWORK: a workflow system supporting rule-
based workflow adaptation. Data & Knowledge Engineering, 2004. 51(2): p. 223-256.
13. O'Brien, P.D. and M.E. Wiegand, Agent based process management: applying intelligent
agents to workflow. The Knowledge Engineering Review, 1998. 13(2): p. 161-174.
14. Shepherdson, J.W., S.G. Thompson, and B.R. Odgers, Decentralised Workflows and
Software Agents. BT Technology Journal, 1999. 17(4): p. 65-71.
15. Wang, S., W. Shen, and Q. Hao, An agent-based Web service workflow model for inter-
enterprise collaboration. Expert Systems with Applications, 2006. 31(4): p. 787-799.
16. Yan, Y., Z. Maamar, and S. Weiming. Integration of workflow and agent technology for
business process management. in The Sixth International Conference on Computer
Supported Cooperative Work in Design. 2001. London, Ont., Canada: IEEE.
17. Zeng, L., et al. An agent-based approach for supporting cross-enterprise workflows. in
Proceedings of the 12th Australasian database conference 2001. Queensland, Australia
IEEE Computer Society.
18. Qu, Y., X. Sheng, and W. Jiao. A Multi-Agent Based Model of Workflow Management. in
10th International Conference on Computer Supported Cooperative Work in Design, 2006.
CSCWD '06. . 2006. Nanjing, China.
19. Hongchen, L. and S. Meilin. Application of agents in workflow management system. in
Fifth Asia-Pacific Conference On Communications and Fourth Optoelectronics and
Communications Conference APCC/OECC '99. 1999. Beijing, China.
20. Qiu, J., C. Wang, and Y. He. Research on application of intelligent agents in the workflow
management system. in 2005 IEEE Networking, Sensing and Control, ICNSC2005. 2005.
Tucson, Arizona, USA: IEEE.
21. Chang, J.W. and C.T. Scott, Agent-based workflow: TRP Support Environment (TSE).
Computer Networks and ISDN Systems, 1996. 28(7-11): p. 1501-1511.
22. Huhns, M.N. and M.P. Singh, Workflow agents. Internet Computing, IEEE, 1998. 2(4): p.
94-96.
23. Merz, M. and W. Lamersdorf, Crossing Organizational Boundaries with Mobile Agents in
Electronic Service Markets. Integrated Computer-Aided Engineering, 1999. 6(2): p. 91 -
104.
24. Alonso, G., et al., Functionality and Limitations of Current Workflow Management
Systems. IEEE Expert, 1997. 12(5).
25. Georgakopoulos, D., M. Hornick, and A. Sheth, An overview of workflow management:
From process modeling to workflow automation infrastructure. Distributed and Parallel
Databases, 1995. 3(2): p. 119-153.
26. Suh, Y.-H., et al., Design of a Mobile Agent-Based Workflow Management System, in
Mobile Agents for Telecommunication Applications: Third International Workshop, MATA
2001, Montreal, Canada, August 14-16, 2001. Proceedings, S. Pierre and R. Glitho,
Editors. 2001, Springer Berlin / Heidelberg. p. 93-102.
27. Lange, D.B. and M. Oshima, Seven good reasons for mobile agents Commun. ACM, 1999.
45(3): p. 88-89.
28. Merz, M., B. Liberman, and W. Lamersdorf, Using Mobile Agents To Support
interorganizational Workflow Management. Applied Artificial Intelligence, 1997. 11(6): p.
551 - 572.
29. Bolcer, G.A. and R.N. Taylor, Advanced Workflow Management Technologies.
SOFTWARE PROCESS—Improvement and Practice, 1998. 4: p. 125–171.
30. Sheth, A. and K.J. Kochut. Workflow applications to research agenda : Scalable and
dynamic work coordination and collaboration systems. in Workflow Management Systems
and Interoperability, NATO Advanced Study Institute on Workflow Management Systems
and Interoperability. 1997. Istanbul, Turkey.
31. Blake, M.B. and H. Gomaa, Object-Oriented Modeling Approaches to Agent-Based
Workflow Services, in Software Engineering for Multi-Agent Systems II, C. Lucena, et al.,
Editors. 2004, Springer Berlin / Heidelberg. p. 111-128.
32. Blake, M.B. and H. Gomaa, Agent-oriented compositional approaches to services-based
cross-organizational workflow. Decision Support Systems, 2005. 40(1): p. 31-50.
33. Kishore, R., H. Zhang, and R. Ramesh, Enterprise integration using the agent paradigm:
foundations of multi-agent-based integrative business information systems. Decision
Support Systems, 2006. 42(1): p. 48-78.
34. Wang, M., H. Wang, and D. Xu, The design of intelligent workflow monitoring with agent
technology. Knowledge-Based Systems, 2005. 18(6): p. 257-266.
35. Yoo, J.-J., et al., Scalable Workflow System Model Based on Mobile Agents, in Intelligent
Agents: Specification, Modeling, and Application : 4th Pacific Rim International Workshop
on Multi-Agents, PRIMA 2001, Taipei, Taiwan, July 28-29, 2001. Proceedings, S.-T. Yuan
and M. Yokoo, Editors. 2001, Springer Berlin / Heidelberg. p. 222-236.
36. Yu, L. and B.F. Schmid. A conceptual framework for agent-oriented and role-based
workflow modeling. in Proceedings of the CaiSE Workshop on Agent Oriented Information
Systems (AOIS99). 1999.
37. Buhler, P., J.M. Vidal, and H. Verhagen. Adaptive Workflow = Web Services + Agents. in
ICWS '03. 2003. Las Vegas, Nevada, USA: CSREA Press.
38. Singh, M.P. Distributed enactment of multiagent workflows: temporal logic for web service
composition. in Second international joint conference on Autonomous agents and
multiagent systems 2003. Melbourne, Australia ACM Press.
39. Buhler, P.A. and J.M. Vidal, Towards Adaptive Workflow Enactment Using Multiagent
Systems. Information Technology and Management, 2005. 6(1): p. 61-87.
40. Joeris, G. Decentralized and Flexible Workflow Enactment Based on Task Coordination
Agents. in 2nd Int’l. Bi-Conference Workshop on Agent-Oriented Information Systems
(AOIS-2000@CAiSE*00). 2000. Stockholm, Sweden: iCue Publishing.
41. Stormer, H. Task Scheduling in Agent-based Workflow. in Int. ICSC Symp. on Multi-Agents
and Mobile Agents in Virtual Organizations and E-Commerce (MAMA’2000). 2000.
Wollogong, Australia.
42. Parunak, H.V.D., Applications of distributed artificial intelligence in industry in
Foundations of distributed artificial intelligence G. O'Hare and N.R. Jennings, Editors.
1996, John Wiley \& Sons, Inc. p. 139-164.
43. Meng, J., S. Helal, and S. Su. An ad-hoc workflow system architecture based on mobile
agents and rule-based processing. in Proceedings of the 2000 international conference on
artificial intelligence (ICAI2000). 2000. Las Vegas.
44. Delias, P., A. Doulamis, and N. Matsatsinis, A Joint Optimization Algorithm for
Dispatching Tasks in Agent-based Workflow Management Systems, in
10th International
Conference on Enterprise Information Systems, ICEIS 2008. 2008: Barcelona, Spain.