From: AAAI Technical Report SS-94-07. Compilation copyright © 1994, AAAI (www.aaai.org). All rights reserved.

TOOLS AND TECHNIQUES FOR ORGANIZATIONALPROCESS DESIGN (Extended AbstracO John Idicula GINTIC Institute of Manufacturing Technology NanyangTechnological University Nanyang Avenue Singapore 2263 idicula@ntu.ac.sg

1. Introduction Organizations need to rapidly adapt to changes in dynamic markets. The design of an organization can hinder or even prevent it from responding quickly to changes in its environment. The processes along with other resources in the organization drive its workand responses to changes in its environment. Processes of an agile organization must be effective, efficient, reusable and well managed. Furthermore, they must ease quick responses to changesin the environmentof the organization. The study of designing organizational processes can stimulate us to design better organizations. It can borrow ideas, tools and techniques fromareas such as artificial intelligence, computer science, coordination science, logic and operations research. Mostof the activities performedin organizations deal with the processing of information to meet various goals. The design of an organization must support the primary activities that help the organization to achieve its goals. Information technology is providing new ways of organizing and performing activities. Transactions that required several hours of human attention can nowbe achieved in a few seconds. There are various kinds of communicationfacilities that have increased efficiency at reduced coordination costs. Information technology based tools are influencing the ways in which

organizations perform their activities. Therefore, we need to design and re-engineer organizations taking advantage of new ways of performingorganizational activities. Twofundamental problems faced in the research area of computational organization design are representing knowledgeand managing coordination activities. An organization achieves its goals by using its knowledge. Much the activity in an organization has little of to do with achieving the organizational goals. Instead, they support those activities that are performed to achieve the organizational goals. Wecan study the problems associated with representing knowledgeand managingcoordination from the perspective of organizational processes. Following a process oriented view leads us to someimportant issues that ought to be addressed in the research area of computational organization design: 1) Howcan we represent organizational knowledge? 2) How can we modeland managecoordination activities in an organization? 3) How we find suitable repcan resentations for knowledgeand coordination in a particular organization? A methodological approach to deal with these questions would be of great practical utility. It would also providea frameworkfor computational organization design and re-engineering. In this paper, we highlight some tools and techniques from various disciplines. These

94

tools and techniques mayenable us to study in detail the issues mentionedabove. Furthermore, they can enrich the study of organizational processes and provide insights for designing organizations better. 2. Organizational Processes An organizational process is a set of tasks that achieveparticular goals of its organization. There are many kinds of processes within an organization. Theyaffect all the work in the organization in some manner or other. Therefore, they deserve a great deal of attention during organization design. Unfortunately, their designs mayremain stagnant throughout the life span of the organization. Analyzing and redesigning the organizational processes mayimprove the performance of work in organizations. Adding the missing essential activities and deleting the present nonessential activities can improve the design of organizational processes. Optimizing the performanceof a process design is a very challenging and interesting task. Usually, it is possible to have morethan one design for a process. It is only natural to choose the best process design to suit particular instances. Therefore, it is essential to have somemeansto differentiate betweenthese process designs using various criteria. Eachprocess has someinherent properties that can be used to evaluate its design. Aset of processes mayalso have someother properties that maybe used for categorizing thembased on their similarities and specialties. 2.1. Complexityof Processes Processes can be classified based on complexity, borrowing results in computational complexity theory. Let us define the size of a process as a function of its inputs. Thenwecan define the complexity of a process as a function of its size. Thecomplexityof a process gives a 95

boundfor the resource requirements of the process for its termination. It is actually a measure of the fundamentalactivities it performs. Some processes are of polynomial complexity while others may be of exponential complexity. As the size of the process grows large, the number of fundamentalactivities performedby a process of exponential complexity grows very fast. We can also say that the processes of polynomial complexityrequire less effort and resources for completion than those of exponential complexity. We can appreciate the notion of the complexity of an organizational process viewing it as a description of somecost associated with the process. For example, the fundamental activities mayinclude the coordination actions the organizational process requires and/or performs to achieve its goal. At this moment,it is not very clear what constitutes a good measure for the complexityof an organizational process. The complexityof a process is an inherent property. No amount of clever design can alter the complexity of a process. Although there is nothing that can change the complexity of a process, there are ingenious meansto managing it. Knowing the complexity of a process is key to optimizing the performanceof a process design in terms of time taken for its execution and other associated costs. 2.2. Similarities and Specialties Similarities amongactivities in organizational processes can form another basis for categorizing them. Usually, there are several types of similar processes in an organization. Similar processes mayalso have special characteristics that differentiate one process from another. For example, we may consider the advertisement process of an organization. The possibilities for advertising a newproduct may include resorting to media such as television,

radio, or newspaper.Thereare activities that are common the processes of advertising in these to media. There are also specialized activities in each of the advertisement processes. Wecan exploit such similarities and specialties of organizational processes to represent, analyze, improve and manage them [Ma93].

way to specify precedence relationships among sub-activities in the process. Malone et al. view coordination processes as ways of managing interdependencies between activities in a process [Ma93]. Using this view, it may possible to characterize difbe ferent types of interdependenciesand to identify appropriate coordination processes for managing them. Then we can specifically indicate in the process representation, howan instance of a particular type of coordination process manages an interdependency. 3.2. Significant Benefits There are several significant benefits in using the organizational process representation approachdescribed in [Ma93].First, it can help to invent new processes easily, if there is an available library of process modelswith various features. Furthermore, rapid and effective reconfiguration of organizational processes is easy for quickly adapting to changes. Second, if there is sufficient information on the organizational processes, redesigning them would be relatively easier than starting from scratch. Changesmadeat higher levels can transfer the properties and attributes downto the morespecialized processes. Third, explicit representation of specialized processes can help to analyze their relative advantages and disadvantages, and to choose an appropriate alternative amongthe various options available. Fourth, the precedence relationships betweenthe sub-activities in a process can help to analyze its correctness, efficiency, effectiveness, as well as timeliness. 4. Process Analysis There are many tools being used for process analyses. The nature of the processes and the analysis affects the choice of the tool. In the following paragraphs wehighlight a tool and a methodfor process analysis. Both of these are widely applicable.

3. Representing Organizational Processes Research in computer science and other disciplines during the last few decades has produced manyapproaches to representing processes. Flow charts, data flow diagrams, state transition diagrams, and directed graphs are some of the commonly used representations. A remarkable and distinct approach of Maloneet al. represents processes at various levels of abstraction, using the concept of inheritance from computer science and ideas from coordination theory about managing dependencies [Ma93]. Theyuse this approach to build an on-line library of organizational processes. This approach allows its users to explicitly represent similarities and differences among related processes. Furthermore, it can also help to analyze and improve existing processes and to design neworganizational processes. 3.1. Inheritance and Specialization Malone et al. make use of the inheritance property as in the object oriented paradigm commonly used now in programming [Ma93]. In the object oriented programmingapproach, increasingly specialized hierarchies contain objects. In contrast, they proposeto develop a hierarchy of increasingly specialized processes. In their approach, generic processes are built from related processes. A generic process decomposes into sub-activities and specializes into morefocused processes. These specialized processes inherit sub-activities and other characteristics of its "parent" process. The specialized processes mayalso add to or change the inherited characteristics. In addition to representing decomposition specialization, there is also a and 96

4.1. Design Structure Matrix Thedesign structure matrix is a tool for representing structural information and relationships of a system [S81a & b, Ep91]. It can be used to capture the flow of information in any system, store the related information, and perform related computations. The rows and columnsof this matrix show the tasks or parameters of the process. Entries in the matrix show the dependencies of a task or parameter upon another. Underlying the information in the matrix is a directed graph. The properties of this directed graph and other information in the matrix can be used for analyzing the process. They can also be used to sequencethe activities in a process and manage other coordination activities a system requires. Viewinga process as a system, we can use the design structure matrix for its analysis. Given a suitable representation of a process, we can modelthe process using the design structure matrix. Entries in this matrix indicate interdependencies between the tasks or parameters involved. If task a does not depend on task b, then the entry at location [a, b] in the matrix takes a zero value. Otherwise, it takes a non-zero value. The design structure matrix and the process representation model of Maloneet al. can complementeach other. Both these representations are suitable for different types of analyses. Wecan perform numerical as well as non-numerical computations, and related analyses on the design structure matrix. A typical application of the design structure matrix is to find a near-optimal sequence for the tasks, subject to various constraints. If the length of each task in the process is known,we could try to sequencethe tasks to minimize the total duration of the process. Consideringthe underlying structure of the matrix, we could find howto group together the agents performing the tasks to minimizecoordination costs while increasing opporttr~ities-for 97

communication between them. Such computations are very useful for managingdesign projects. Recent works using design structure matrices showthat they are very useful in studying instances of design processes. 4.2. Correctness of Processes Oncethe activities in a process are sequenced, we can test the correctness of the design of the process [Ha87]. Proving correctness of a process design is not always easy. It involves two steps, namely proving partial correctness and termination. Its basic element is choosing checkpoints in the process design. Checkpointsconsist of start and stop points, and sufficiently many intermediate locations, so that each loop contains at least one checkpoint. Assertions can be attached to various intermediate checkpoints and their invariance can be confirmed by proving their local properties. Termination of processes can be proved by using checkpoints and showing convergence of some quantity dependingon the process variables and representations. Determining the fight checkpoints and proving correctness of process design require careful attention. 5. Summary In this paper, we briefly highlighted sometools and techniques that, weconsider, are of importance to the study of designing organizational processes. They provide a firm foundation for designingbetter organizations. We speculate that the knowledge of process complexity is useful for developing better organizational process designs. Wecan draw upon results in computational complexity theory for this purpose. In computational complexity theory, the common trend is to consider resource requirements such as time and space. In the study of organizational processes, we believe that it would be beneficial to consider other factors such as coordination costs and related communication costs.

Wedo not know at this point in time what constitutes a good measure of complexity for the organizational processes. The process complexityshould serve as a universal basis for comparing processes. The kind of the technologies used for managing communicationand coordination of a process should not alter its complexity. The formal techniques applied in computer program development are useful to analyze the correctness of organizational processes. Those methodscan also help to improve the efficiency and effectiveness of each task in a process. The design structure matrix can help to manageinterdependencies between the tasks in a process. Wehave found it to be very useful in studying design processes. The approach of Maloneet al. provides an elegant representation for organizational processes. This approach is being used to develop an on-line handbook for organizational process. The process representation approach of Malone al. and the design structure matrix are et very useful for representing knowledgeon organizational processes and managingcoordination. Weneed to develop a science base for studying organizational design. A systematic approach to designing better organizations should be of great practical utility. Thestudy of designing better organizations will benefit from further developmentsin representing knowledge and managingcoordination. References [Ep91] Eppinger, S.D. "Model-Based Approaches to ManagingConcurrent Engineering," International Conference on Engineering Design, Zurich, August1991.

[Ha87] Harel, D. Algorithmics : The Spirit of Computing, Addison-Wesley, Reading, MA, 1987. [Ma93] Malone, T.W., Crowston, K., Lee, J., and Pentland, B. "Tools for Inventing Organizations: Toward a Handbook of Organizational Processes," 2nd IEEE Workshop on Enabling Technologies Infrastructure for Collaborative Enterprises, 1993. [S81a] Steward, D.V. Systems Analysis and Management:Structure, Strategy and Design, Petrocelli/McGraw-Hill, NY, 1981. [S81b] Steward, D.V. "The Design Structure Matrix: A Method for Managingthe Design of Complex Systems," IEEE Trans. Engineering Management, August 1981.

98