Discussion

This case illustrates a number of advantages of using a specialization hierarchy in combination with the explicit representation of coordination and dependencies. First, this field test showed that specialization can substantially reduce the amount of work necessary to analyze a new process. By simply identifying a process as a "hiring process", for example, a great deal of information can be automatically inherited. Then, only the changes that matter for the purpose at hand need to be explicitly entered. This helps support a rapid assessment of the basic features of a process, rather than laborious detailing (what Hammer and Champy, 1993, refer to as "analysis paralysis"). For example in the field test, the team chose to ignore nearly all of the "as is" analysis that had previously been done by Firm A and focus on a very simple, abstract view of the hiring process and its first-level subactivities. This level of detail, alone, was sufficient to generate all the insights described above.

Second, the specialization hierarchy provided a powerful framework for generating new process ideas. For example, some of today's "best practice" databases support cross-fertilization across industries within the same business function, but we do not know of any others that would support the kind of cross-fertilization across business functions (from purchasing to human resources) described above.

Since coordination processes are often those most susceptible to being changed by information technology, a particularly important use of this approach is to use generic knowledge about alternative coordination mechanisms to generate new process ideas. For instance, the ideas about using bidding to allocate recruiter time were stimulated by very generic knowledge about coordination, and would presumably be more feasible because of the cheaper communication made possible by information technologies (see Crowston, 1997, for other similar examples).

Another feature of our approach that makes it particularly useful for generating new process ideas is that we focus attention on processes as distinct entities that can be described independently of organizational structures or the roles of particular people or groups. This "process-oriented" approach to business seems particularly useful, in (a) identifying new ways of doing old tasks, even if the new ways involve very different actors and (b) managing connected processes that span organizational boundaries, either across groups in a single firm or across firms in "networked" and "virtual" organizations.

In addition to these advantages, our process-oriented approach has limitations, too. For instance, any static process representation can give the impression that the process is more stable and routine than most business processes actually are. In contrast to most other process representations, however, our approach helps us explicitly deal with this issue by representing the stable—or typical—aspects of a process at the generic level and then also representing as many specialized variations as is useful.

Another risk of having libraries of explicit process representations like ours is that people will interpret them too rigidly. While it is sometimes appropriate to collect prescriptive rules or procedures in a handbook like ours, we think that in most situations a Process Handbook will be most useful as a resource to help people figure out what to do, rather than as a prescription of what they should do.

The Editorial Challenge One of the most important ways in which our approach differs from many other computational approaches to similar problems is that we do not rely primarily on intelligent computer systems to analyze, reason about, or simulate business processes. Instead, we place substantial importance on the role of intelligent human "editors" to select, refine, and structure the knowledge represented in the handbook. This approach has both strengths and weaknesses.

On the one hand, it allows us to take advantage of human abilities to analyze, organize, and communicate knowledge in ways that go far beyond the capabilities of today's computers. For example, the task of developing good generic models for the marketing and sales process is similar, in many ways, to writing a good textbook or developing comprehensive theories about marketing and sales. Human abilities to do tasks like these will almost certainly exceed those of computers for the foreseeable future.

On the other hand, relying on human effort in this way means that the success of our approach depends significantly on the quality and amount of human intelligence applied to the problem of generating and organizing knowledge in the system. For example, a complex and confusing network of poorly organized process categories may be even worse than no categories at all.

In general, as process descriptions are added to the handbook, we will face a problem that is analogous to that faced by researchers in many fields: how to ensure that results cumulate in a meaningful way. Since we foresee a wide variety of potential users and contributors, it would be unrealistic to expect equal rigor from all of them. Rather than attempting to enforce uniform standards, we plan to allow a wide variety of data from diverse sources, but to require that the specific sources, methods, and significance of that data be described in enough detail to allow users of the handbook to judge whether it is valid and reliable enough for their own purposes. In this respect, the handbook has an advantage over more formal approaches because it allows many alternatives to co-exist in the system. At the same time, this openness contributes to the editorial problem of ensuring that the entries are consistently and usefully classified. We believe that adopting solutions analogous to those that have already been found successful in other domains is a promising approach. For example, we have found it useful to think about roles like authors, editors, and reviewers for groups of entries in the Process Handbook.

It is also encouraging to note that the specialization structure of the handbook provides a potentially powerful advantage that has not been widely available to any knowledge-generating communities before: Well-organized and accurate process knowledge at the "left" of the specialization network is automatically inherited throughout the other parts of the network where it applies. In this sense, then, the system amplifies the effort of intelligent humans by automatically linking their work to a variety of contexts where it may be useful.