A firm decided to redesign its research and development process. Because the effort was critical to its success, the firm applied two parallel approaches to the process. One was a classical reengineering effort in which a small group of managers and consultants designed a radically different way to do research. In the other approach, a small team of researchers began to collect the organization’s knowledge about the entire R&D process, regulatory approvals, and international variations. The knowledge was embedded in a series of documents and in a Weblike hypertext computer system. The company also hired several “corporate anthropologists” to study how researchers used the knowledge.

A year after the two efforts began, the reengineering effort was stalled. The new process vision was viewed as too ambitious at a time when many other aspects of the company were changing. Scientists muttered that no one could force them to practice “bad science.” The more subtle knowledge collection effort, however, seemed to be succeeding. Many researchers had reviewed the documents and suggested changes or additions to the knowledge base. A greater sense of cross-functional understanding seemed to prevail; regulators, for example, reported that scientists were more willing to supply them with early research information. Specific improvement levels, however, were difficult to measure with confidence.

This example illustrates how the classical top-down reengineering approach for improving administrative or operational work is often insufficiently participative or flexible for improving work by autonomous knowledge workers. In a study of knowledge work improvement projects in thirty organizations, we have found that although there are substantial benefits from viewing knowledge work from a process perspective, there are significant differences in how process concepts and methods are applied to knowledge work versus operational or administrative work.

In 1978, Peter Drucker wrote, “To make knowledge work productive will be the great management task of this century, just as to make manual work productive was the great management task of the last century.”1 By 2004, professionals and managers are projected to account for 25 percent of all U.S. jobs.2 More of an organization’s core competencies will center around managing knowledge and knowledge workers. Industrial growth and productivity gains will depend heavily on improvements in knowledge work.

Traditionally, organizations have tended either to ignore knowledge work improvement or to manage it in a hands-off manner. For example, a recent survey found that of some 435 organizations with reengineering initiatives, only 4 percent had redesigned, for example, the product development process, a knowledge work process that is critical to long-term performance.3 The importance of knowledge work to today’s organizations, however, suggests that the current laissez-faire approach to managing knowledge work and knowledge workers will not suffice.4 A viable approach is critically needed for improving knowledge work. In this paper, we describe how thirty organizations have attempted to improve knowledge work processes and what seems to be working for them.

We begin by defining knowledge work. Then we note the challenges that it presents to a process orientation. With knowledge processes understood in a modified light, we discuss how our research sites have attempted to improve knowledge work processes and which approaches seem to work under what circumstances. Finally, we present a framework of redesign strategies for improving knowledge work processes.

What Is Knowledge Work?

Both knowledge and knowledge work are difficult to define precisely. Quinn et al. equate knowledge with professional intellect.5 To them, professional intellect in organizations centers around know-what, know-how, know-why, and self-motivated creativity. Others define knowledge more narrowly as agreed-on explicit or formal facts, rules, policies, and procedures, whereas they see skills as competencies that can generate explicit knowledge.6 Skills are learned by doing; knowledge is learned by studying or investigating.7 Nonaka defines knowledge as “justified true belief,” where beliefs are dynamic, relative, unstable, and person dependent. He distinguishes between “tacit” knowledge that is personal (i.e., knowledge not easily expressed and communicated) and “explicit” knowledge that can be codified and expressed in formal language.8 In contrast to “knowledge,” relatively few writers have attempted to define “knowledge work.” Davis et al. define it as “a set of activities using individual and external knowledge to produce outputs characterized by information content.”9

In our definition, knowledge work’s primary activity is the acquisition, creation, packaging, or application of knowledge. Characterized by variety and exception rather than routine, it is performed by professional or technical workers with a high level of skill and expertise. Knowledge work processes include such activities as research and product development, advertising, education, and professional services like law, accounting, and consulting. We also include management processes such as strategy and planning.10

Many other types of jobs involve the use or application of knowledge; one restaurant firm views its counter personnel as knowledge workers. Leonard-Barton has also described the benefits of viewing traditionally menial or repetitive jobs from the standpoint of their potential for knowledge creation and acquisition at Chaparral Steel.11 We agree that workers in these types of jobs can make significant knowledge contributions. However, such jobs do not have knowledge activities as a primary component of their roles, and they are not professional or technical workers with high expertise. While any definition is somewhat arbitrary, we believe ours is consistent with the popular understanding of knowledge work.

Applying a Process View to Knowledge Work

A process approach to knowledge work attempts to separate knowledge work — at least to some degree — from the idiosyncrasies of a particular knowledge worker. In the past, the focus was on managing knowledge workers rather than knowledge work. Managers treated the way that knowledge workers performed their activities, and sometimes even the time, cost, and quality with which knowledge work outputs were produced, as an impenetrable “black box.” Any sense of management often ended when the employee entered or received tenure in an organization.

We believe that firms can do better than the current black box approach by applying a process approach. The process approach allows an end-to-end view of how best to structure, sequence, and measure work activities to reach targeted outcomes. Processes, as one of us wrote, are “a specific ordering of work activities across time and place, with a beginning, an end, and clearly identified inputs and outputs: a structure for action.”12 The process approach promotes an examination of what and how things are done from a viewpoint of producing value for a customer. Although substantial benefits may be derived from a process approach, there are also significant challenges to its application (see Table 1).

Knowledge work is untidy. Unlike operational or administrative business processes, where tangible inputs are acted on in some predictable, structured way and converted into outputs, the inputs and outputs of knowledge work — ideas, interruptions, inspirations, and so on —are often less tangible and discrete. There are no predetermined task sequences that, if executed, guarantee the desired outcome. Knowledge workers may operate by an intuitive feel for how to accomplish their work or through accumulated experience.

It is also difficult to separate a knowledge process from its outcomes. In high-tech organizations, breakthrough products often require the invention of a new process. The product manager has nearly total freedom in the development process. The lack of separation between process and outputs is also reflected in the way outputs are valued. The more laborious the process or expensive the input (i.e., expert knowledge worker), the more value placed on the outcome.

A process approach assumes customers and measurements that relate to value generated for the customer; quantitative measures in administrative areas rely heavily on productivity assessments or transformations of tangible inputs to physical outputs. In service or information-intensive sectors such as financial services and education (some of which is knowledge work), work productivity is measured on the basis of input rather than on the effectiveness of the transformation process.13 Knowledge work shares similar difficulties in measurement with service processes.

A process approach also assumes some commonality of activities and outcomes among individuals. Yet knowledge workers tend to enjoy high levels of freedom in how and when they perform work activities. There is good reason for this autonomy; much knowledge work is professional in nature, i.e., some governing or organizing body trains and accredits practitioners, defines performance standards, and monitors compliance with standards. Mintzberg observes that an organization hires “duly trained and indoctrinated specialists — professionals — for the operating core and then gives them considerable control over their own work.”14 Managing professionals, therefore, is a different task from managing administrative or operational workers, requiring a manager to cede day-to-day task control to the professional worker while maintaining control and direction over strategic issues.15

Knowledge workers are likely to resist standard routines; in fact, the level of discretion and autonomy often separates knowledge workers from administrative workers. Some studies have reported that the job characteristic engineers and managers care about most is autonomy. The autonomy in turn leads to differences in work outcomes. The best information system developers, for example, can produce tenfold the amount of programming as the worst. As a result, and in contrast to typical operational processes, there are few standards for efficient allocation of employee time for knowledge work. The rise of “virtual offices” and managers working at all hours only exacerbates this issue.

Information technology often makes new process designs possible in operational and administrative areas. The abstract and unstructured inputs to and outputs from knowledge work processes, however, make the application of technology more difficult. As work becomes more knowledge intensive, rapid manipulation of data across distances has less impact; “richer,” more face-to-face communications are more important. Technology can support knowledge work processes, but it must be implemented with sensitivity to the nature of the work and its practitioners.

Toward a New Conception of “Process”

Given these challenges, the process orientation of knowledge work is different from that of operational or administrative work. Ever since Frederick W. Taylor’s research, a process orientation has focused on the breakdown of work into small, standardized, measurable tasks — a “work breakdown structure” approach. In knowledge work, however, the nature of the activity and the people who perform it resist heavily structured, standardized approaches.16

Therefore, a modified process approach that accommodates these challenges is needed. Knowledge work redesign might, for example, alter the space and context in which the work is carried out without explicitly addressing the work flow. When Chrysler rethought its vehicle design process, among the design principles were adopting stretch goals, moving product developers into headquarters, using teams, having a common budget, linking individual performance to the overall vehicle success, and establishing a “platform team” that brings together engineers, manufacturers, accountants, and so on. The detailed activities and steps in new vehicle development were left to each new car development team.17

Methods and Approaches to Improve Knowledge Work

To develop a workable approach for improving knowledge work, we studied improvement projects in thirty organizations (see Table 2). In semistructured interviews, we talked with either the owners or the designers (or both) of each process improvement effort. We also visited seven organizations for more in-depth interviewing. We analyzed the projects’ objectives, methods, and design scenarios. In contrasting our findings to our experiences with administrative and operational work-flow redesigns, we discovered that the projects varied by their primary orientation to knowledge. We distinguished five different primary orientations to knowledge among the thirty projects: acquisition, creation, packaging, application, and reuse of knowledge.

  1. Some processes consisted of finding existing knowledge — understanding knowledge requirements, searching for it among multiple sources, and passing it along to the requester or user. An example is a competitive intelligence process in an insurance company.
  2. Other processes involved creating new knowledge. Examples are the research activities in a pharmaceutical firm and the creative processes in advertising, writing books or articles, or developing a movie.
  3. Knowledge work processes can package or assemble knowledge created externally to the process. Publishing is a prime example of knowledge packaging. Even though it does not create new knowledge, the editing, design, and proofing processes qualify as knowledge work.
  4. Certain processes primarily apply or use existing knowledge. In these processes, the creation of new bodies of knowledge might be actively discouraged. For example, in the redesign of an auditing process, the auditor was expected not to create new knowledge about financial reporting but to interpret and apply existing procedures to a company’s financial transactions. Similarly, a physician is not normally expected to experiment on a patient to create new knowledge but rather to apply existing medical knowledge.
  5. Some firms have a primary focus on the reuse of knowledge. They promote learning but focus on separating it from prior knowledge and leveraging that prior knowledge as much as possible. We noticed reuse in product development processes, in which engineers were encouraged to reuse existing components, and in software development processes. Indeed, reuse is a core strategy for information systems developers adopting object-oriented technology.

The different knowledge orientations were in turn associated with different objectives and methods for knowledge work improvement. For example, in knowledge creation, explicit improvement objectives were rarely stated. Projects focused on creating an environment of spontaneity and autonomy. Any process-oriented management focused on education and reward of outputs rather than on adherence to a series of steps or activities. Publishers left authors of books, for example, to their own devices for how they carried out their process but paid the bulk of their compensation after they finished their books. Pharmaceutical companies encouraged researchers to review a compilation of expert knowledge on the new product development process but did not monitor them on their adherence to process steps.

In contrast, a key issue in the application of knowledge was disseminating established knowledge to the front lines for use. Objectives were clearer in terms of cycle time, cost, and quality. Processes for using information were defined, as in health care, with defined protocols for patient care. Or they were supervised closely, e.g., through cross-functional teams involving both knowledge workers and front-line workers.

Improvement Objectives

Along with the differences in knowledge orientations in the thirty initiatives we studied, we found significant deviations between these projects and what we had observed previously with successful administrative and operational reengineering projects. In our reengineering work, we have advocated setting explicit, measurable, and highly ambitious objectives in order to achieve breakthrough results.18 Firms should strive for order-of-magnitude improvements such as “10x.” A broad survey of reengineering initiatives also suggested that radical goals were more likely to be achieved with explicit, radical change objectives.19

We discovered in our research that the objectives for knowledge work improvement are less ambitious. Not only did we find few projects with radical change objectives, but we also found no specific change goals at all for most projects. When asked the objectives of the project in which they were engaged, managers frequently stated, “We want to reduce the time to edit and publish a book” or “We want to significantly reduce the cycle time to approve a capital project.” In the latter example, the project manager rejected a consultant’s advice that more explicit objectives would be more likely to produce change. Given the fact that many of the knowledge workers did not work for him and had typically autonomous jobs, he did not feel comfortable creating an explicit target.

We did find a few projects with quantified cost or cycle-time reduction goals. For example, two pharmaceutical firms were attempting to shorten their drug development cycle time. One firm was quite ambitious, striving for a cycle of six to seven years instead of the current ten to twelve years. The other firm wanted a reduction of only three to six months in a nine- or ten-year cycle (again, note the imprecise objectives). Another firm that wanted to reduce cycle time was a computer company focusing on the configuration subprocess within the order management process. It wanted to move from a five-day average cycle time to a virtually instantaneous configuration using a knowledge base. (Configuration is a knowledge work subprocess within a generally administrative process.)

A product development organization was attempting to reduce the cycle time for its new product approval and sign-off process by segmenting a knowledge work process into its subcomponents. Analysis revealed that many people involved in the approval cycle did not need to approve the product design but only to know about it. By separating information distribution from approval and clarifying participants’ roles, the company cut the number of required signatures by two-thirds and is nearing its goal of reducing the average cycle time for product sign-off by 70 percent.

Another segmentation-oriented objective was the goal of freeing workers to do knowledge work by reducing their administrative tasks. A pharmaceutical firm’s goal in improving the research process was to free scientists from time-consuming, bureaucratic steps such as record-keeping and document production. In two human resource (HR) processes we observed, the goal of an improvement initiative was to relieve HR knowledge workers from the burden of answering routine employee questions about benefits, work-life options, and job change issues. In one firm, the process change involved centralizing responses to such questions to gain economies of scale; in another, answers to typical questions were put into an easy-to-use textual database.

Other objectives for knowledge work improvement included:

  • Make an implicit knowledge process explicit and consistent.
  • Add knowledge to a process to add value to the process customer.
  • Involve the customer in the process to increase satisfaction with results.
  • Share knowledge more effectively throughout the process.
  • Improve the execution of programs and initiatives.

The breadth of objectives for knowledge work initiatives may be a bellwether for other types of reengineering projects. Knowledge work process improvements often focus on increasing value and making products and services more desirable in the marketplace. Reengineering has been criticized for focusing on cost reduction and not focusing enough on increasing revenues.20

A Continuum of Improvement Methods and Approaches

The improvement methods for knowledge work processes also deviated from those often associated with administrative or operational reengineering. They deviated from the traditional black box approach for managing professionals as well.21

Hammer and Champy’s aggressive methods are at one end of a continuum for improving knowledge processes.22 At the other extreme are traditional “laissez-faire” views of knowledge work, in which knowledge workers are fully responsible for designing and executing their own work, and any knowledge work process is seen as a black box.23 Information systems managers at Hewlett-Packard referred to these approaches as “process versus heroes” and conducted an extended electronic debate on the merits of these polar approaches for the software development process.24 (See Table 3 for the contrasting improvement methods.)

Neither extreme seems appropriate for most companies’ knowledge work environments. The laissez-faire approach might be characterized as finding good knowledge workers and leaving them to their own devices, only to measure how quickly and how well they produce outputs. Archetypal examples of this philosophy are the pharmaceutical scientist who experiments with new compounds in the lab for years before identifying a specific drug, or the university faculty member who has several years to become a productive researcher before being evaluated for tenure and promotion. In both examples, the unit being evaluated is the individual rather than the larger organizational unit.

The laissez-faire view advocates a hands-off approach to managing change. Since knowledge workers are autonomous, they do not assume change unless it is their own idea. If individuals do adopt the work designs of others, it is not because they are compelled to do so, but because they have been persuaded intellectually of the benefits. Because firms see workers as intelligent, they pay substantial attention to understanding how they do their work before recommending changes. They understand that there must be a reason why the process is as it is.

In the reengineering approach, by contrast, firms strive to change how knowledge workers do their work through the redesign of day-to-day activities. Firms decompose work activities into microlevel steps. They assess performance on a daily or even hourly basis. In this approach, evaluation may be at the level of the cross-functional process team; the individual’s identity is subsumed within a larger group. The reengineering approach assumes that change is managed from the top down. In fact, Hammer and Champy argue that “it is axiomatic that reengineering never, ever happens from the bottom up.”25 Workers comply with new work designs because they are mandated to do so. The “as is” process is deemphasized, and most time is spent on the “to be” work design. Since the existing process is not worth saving in this approach, and “out of the box” thinking is highly desirable, it makes sense to employ mostly outsiders — either to the process or to the organization — to do the redesign.

Finally, the greatest barrier to change in the laissez-faire view differs from that of the reengineering view. It is common to observe fear of change on the part of workers whose jobs will be “reengineered.” In the laissez-faire context, such fear may not be assumed; in fact, knowledge workers may be stimulated by change of their own making. The greatest barrier to improvement we observed was, rather, strong loyalty to a profession or discipline. In product development processes, for example, the barriers arising from disciplinary loyalty to engineering or manufacturing professions are well documented.26 In pharmaceutical research, scientists insist on completing research to the professional standards of their discipline before sharing information about it throughout the process. To release findings prematurely, one scientist told us, would be bad science —even if it meant great savings for process steps further down the line.

The Research Sites along the Continuum

What seemed to work for most of the projects we studied was an improvement method that came near the middle of the continuum. The projects employed some aspects of the laissez-faire approach and some components of reengineer-ing.27 The most common approach was a participative one, which involved knowledge worker participation but not full control of the change process. Knowledge work improvement projects, we found, tended to be quite broad in terms of participation in the work design. On several projects, the design teams included between fifteen and twenty employees, roughly double the average size of design teams we have observed in administrative and operational reengineering projects.

Our respondents cited the need for knowledge worker participation in order to secure their help in implementing a new process. Also, since knowledge work frequently involves expertise that is fragmented among multiple workers, in order to understand the process fully, more workers needed to participate. For example, in the university fund-raising project we studied, individuals specialized in either individual, government, or foundation gifts. To design a better process, a representative from each type of fundraising needed to be on the team.28

The focus of analysis for knowledge work projects was less oriented to activities and tasks and more to process outputs. Projects addressed what kinds of documents are or should be produced by the process, when a particular output should be produced, how much it should cost to produce it, and so on. In three projects we studied, process improvement consisted largely of eliminating outputs that customers no longer deemed useful or worth the cost.

Similarly, many firms viewed the commitment to implementing new knowledge work designs as persuasion rather than mandate. They offered, marketed, or communicated the new designs through education rather than forcing them on those who performed the knowledge work. This may be viewed as either participatory management or a power struggle; in knowledge work, power is more likely to be held by those with the critical knowledge, i.e., knowledge workers rather than managers.

An international bank, for example, developed a new lending process using a geographic region as the primary design site. After it completed and tested the design, the project team leaders produced documents, arranged one-on-one meetings, and held education sessions to try to persuade other regions to adopt aspects of the new design. Bank managers felt that a more heavy-handed approach would run counter to its culture of autonomous knowledge work.

Most projects we studied involved fewer outsiders (such as consultants) than we had observed in classical reengineering projects. We also saw outsiders with softer analysis methods, such as ethnographic studies. By this, we mean intervention techniques involving analysts’ participation in the knowledge work, usually over a long period of time and involving detailed description of the broad work environment and cultural and behavioral issues.29 In short, rather than a consultant interviewing workers over a brief period, in ethnography, anthropologists observe the workers under analysis for several months. These methods involve systematic observation of work approaches, sometimes with detailed analysis of videotapes to understand work activity.

We observed this approach at two firms. A pharmaceutical company used ethnographers to study how a new knowledge base was used in the drug development process. The goal was to learn how this intervention was working within the drug development team. The method showed promising results, although no drugs developed with its help have yet come to market.

The other firm, a high-technology manufacturer, used ethnographers to study its customer service process. The firm was attempting to shift repair calls from field service personnel to telephone service workers. While aspects of the telephone support were purely administrative, the workers did have a simple knowledge base at their disposal to diagnose customers’ problems. The ethnographers realized that some telephone service personnel, if given enough opportunities to acquire knowledge of products and repair procedures, could lead customers through a repair procedure over the phone, saving hundreds of dollars on each service call.

Ethnographic methods seem to fit well with knowledge work. The analyst can understand the many factors in how work is done. The subjects of the study, i.e., the knowledge workers, are not insulted by the assumption that all their work can be learned in a brief interview. However, one potential problem is that ethnographers do not want to generalize beyond their specific observations. A balance must be struck between paying sufficient attention to context on the one hand and making timely recommendations and implementing improvements on the other.

Locating a Project along the Continuum

What determines where a firm should locate a project along the knowledge improvement method continuum? Although most of our research sites had adopted a position near the middle, we found enough deviations to develop tentative guidelines. The appropriate improvement method depends on the primary orientation of the work, the culture of the organization, and the time and level of risk that the project can bear.

For example, organizations attempting to improve knowledge creation processes should steer toward the laissez-faire end of the continuum. These knowledge workers are perhaps the most autonomous and most resistant to microlevel activity design. By virtue of their education and creativity, they may also be the best qualified to design their own work.

At the other extreme, firms wanting to reduce or eliminate the amount of administrative activity in knowledge work processes might employ change approaches closer to the reengineering end of the spectrum. Many knowledge workers do not like such work anyway and might submit to change programs designed to reduce it.

Most other knowledge work segments, e.g., those focused on finding, packaging, applying, or reusing knowledge, are well suited to participative approaches in the middle of the continuum. Of course, a firm needs to choose the specific methods and tactics used for such processes carefully; individual elements of a change program may approach either extreme.

The organizational culture also dictates the position adopted for knowledge work process change. Organizations with long histories of leaving knowledge workers alone should not move to the other extreme without strong provocation. In those establishments where knowledge workers play significant roles in governance (e.g., universities and hospitals), it may be necessary to take a laissez-faire approach. On the other hand, when an entire organization is undergoing reengineering, it is more feasible to adopt that style of change approach for knowledge work processes.

Finally, the approach adopted depends on the time and risk parameters for the project. Laissez-faire approaches take longer to implement but involve lower risk of obvious failure — the departure of several key knowledge workers, for example. Reengineering approaches are more visibly risky but can be fully implemented within a year or two. There may also be a cost dimension in the choice of change approaches; laissez-faire approaches are generally less expensive because they use insiders and need lower levels of detailed process analysis.

Redesign Strategies for Knowledge Work Processes

We have emphasized that in knowledge work redesign, the decisions about detailed work flow must often be left to individual knowledge workers. Hence, the primary strategies for changing knowledge work lie in three areas:

  1. Firms can change knowledge itself by reducing (or, in some cases, creating) a unit of knowledge that workers can reuse or access or by improving knowledge capture techniques.
  2. Firms can improve knowledge work by changing the physical location of where and with whom people work. This change typically involves collocation, new or modified team structures, or new roles. (These first two strategies correspond best to the laissez-faire end of the change approach continuum; they leave decisions about the flow of the work to those who perform it.)
  3. Firms can use technology to bolster knowledge work by, among other things, creating knowledge bases and enabling telecommunications infrastructure and applications. Because they heavily influence process flow, technology changes are more consistent with reengineering-oriented approaches to knowledge work, and most of the projects in which we found them were being done in that context.

Changing the Unit of Knowledge

Companies often want to make knowledge more portable, modular, accessible, and recordable; in short, they attempt to make knowledge easier to manage as a discrete object. For example, one large publisher is attempting to make the unit of knowledge the “idea” (as expressed in a piece of text or a graphic representation) instead of the book. Stored electronically, different authors’ idea modules can be combined and recombined in customized configurations, enabling college professors, for example, to construct a textbook to precisely meet their course requirements. The same content can also be used across multiple media, e.g., magazine article, book, on-line database, and CD-ROM. While there are some legal issues, the publisher is convinced that by making the unit of knowledge smaller, it can be reused and leveraged to a much greater degree.

In a high-technology manufacturing firm, a project was underway to restructure, modularize, and index product documentation. Product developers and marketers spent inordinate amounts of time searching for and recreating product specification documents, design documents, and regulatory requirements documents. In many cases, documents were redundant or even contradictory. The company’s goal was to create standard and uniform product documentation to be stored electronically. Documents would be stored at the lowest component level so that information on more complex products could be assembled rather than rewritten. After a transition period, management would accept only electronic submissions to ensure compliance.

A pharmaceutical firm focused on the preparation of the New Drug Application (NDA) as a knowledge work process. Traditionally, the NDA had been a large, somewhat monolithic entity. However, the company needed to assemble applications differently for different markets. It also needed to make changes quickly in various sections of the NDA to satisfy regulators. The company now creates modular sections of an application rather than one integrated document. It can assemble the modules in different combinations to satisfy the requirements of different regulatory bodies. As a side benefit, the company found that regulators were able to evaluate its submissions more easily because of their modular structure.

Firms also addressed knowledge work improvement by changing the way they captured and stored knowledge. The pharmaceutical manufacturer that developed a “book” of process knowledge and best practices in new drug development employed this strategy. It comprehensively indexed and cross-referenced the document and its electronic equivalent, so that a scientist could see how the information he or she created was employed throughout the entire process. Again, the goal was to reduce cycle time for drug approval and to increase the quality and integrity of the research on a particular drug. In this way, the tacit, local knowledge of scientists became explicit, widely disseminated knowledge of the organization.

A large utility company adopted a similar but more automated approach to knowledge capture and storage. It undertook a project to capture the knowledge of its managers, financial analysts, and engineers about how to analyze, implement, and manage capital projects. The project’s ultimate goal was to create a computer-based expert system; in the short run, the knowledge was captured in a series of documents. The company wanted not only to streamline its own activities when facing a capital project but also to sell this knowledge to other utilities facing similar situations.

Changing Where and with Whom People Work

Frequently, the simple step of putting people together to work in the same room greatly enhances knowledge work effectiveness. In information systems development processes, significant gains have been achieved when IS professionals have worked side by side with users. In a small, informal survey of ten firms’ efforts to improve systems development processes during the past five years, the firms rated the location changes as the most effective of any intervention — over changes in methodology, tools, roles, or metrics.30

In a large oil company we studied, the process for evaluating whether and how much to bid on an oil lease was redesigned with the objectives of reducing cycle time and improving evaluation quality. In the old process, geologists and geophysicists would separately perform exhaustive geological analyses. These analyses, sometimes inconsistent with each other, would be passed to groups of facilities engineers, drilling engineers, platform engineers, and pipeline engineers. Each group would perform its own separate analysis of the cost to extract the oil and/or gas. Business planners would then determine what sort of return could be achieved from the lease and what amount, if any, to bid for the site. The typical evaluation could take up to a year.

However, the firm faced a three-month deadline to respond to a potentially lucrative lease opportunity. To accomplish this, geologists had to work in one room with geophysicists, facilities engineers — drilling, platform, pipeline, and so on — and business planners. When questions or issues arose, they were immediately resolved. While the facilities or business planning group was performing its analysis, the geology team would be taking a second pass at its initial analysis. The company submitted a winning bid within the allotted time frame. While it will be years before oil is pumped from the ground, the company sees the acquisition as a success; it has been approached by two other oil companies to partner in the site development.

Often, shared physical experience is essential for a shared view of the process. At National Semiconductor, there is frequently a need to transfer a particular type of “fab” or fabrication facility from one location to another. For example, the company may want to duplicate a fab based in Santa Clara, California, at its Scotland manufacturing site. Managers have discovered the one best predictor of successful transfer of a fab: the time the receiving group spends physically together with the transferring group. This finding has been tested during several different transfers.

We also found instances when employees from different business functions — who may have already been in the same location — were combined into teams. In addition to joint location, the team members then shared a common purpose, common measures, and often cross-training on tasks and skills. Perhaps most important, the knowledge work process was more coordinated, simply because it was the team’s job to complete it.

At a fast-food firm, a primary objective in improving the marketing process was to better connect the design of marketing promotions with their execution. The headquarters personnel usually developed promotions, and franchises in the field executed them. In the new process, a cross-functional team of headquarters marketers, field managers, and representative franchise supervisors designed promotions. Early results indicate that the franchises are more consistently implementing the promotions.

In a university fund-raising process, the development office had historically managed fund-raising. Other administrators, faculty, and even students sometimes had contact with prospects, but these contacts were seldom coordinated or even reported to the development office. The redesigned process organized the different groups into teams, each with targeted prospects. In effect, each prospect had an “account team.” The team approach has been combined with better prospect segmentation and is already yielding a higher level of contributions. Eventually, the teams will also have a prospect information system to review contact and contribution histories and record new contacts.

A pharmaceutical firm included not only researchers on the improvement team but also an information specialist. The specialist’s role was to provide information sources dealing with both research and market acceptance issues for the new drugs under development. The managers have observed that only the direct membership of information providers on the team seems to get researchers to focus on the desired information.

Employing Technological Enablers

Information technology tools such as discussion databases, knowledge bases, and specialized telecommunications are employed to improve knowledge. Typically, technology is applied in concert with another intervention approach. For example, a consulting firm has developed a system for storing knowledge that consultants can add to or access. The system also tracks document retrieval rates, allowing monitoring of which contributions are most frequently sought. But the system would not be as successful as it is without the related changes in the firm’s incentive and reward schemes. These encourage contributions to the knowledge base — measured in part by the number of times an author’s document was accessed — as a criterion for promotion. Many other knowledge work processes we studied involved such combinations, including the capital projects process mentioned earlier.

We found only one incidence of an expert system, the preferred approach to capturing and formalizing knowledge in the 1980s. The managers of several firms we interviewed noted that they had tried expert systems in the past but found them too difficult to maintain over time. We did find several instances of firms wanting to implement broad knowledge bases in their organizations — for purposes of configuration, customer support, capital budgeting, and audit. Most of these knowledge bases are not yet implemented. The firms generally underestimated the difficulties of structuring and maintaining a knowledge base. Few members of process design teams who recommend such knowledge bases have experience in creating or implementing them. These approaches assume that the knowledge an organization uses is explicit and structured; this may not be the case, however. Knowledge used in knowledge work is often tacit.31

Some companies in the study also implemented more traditional databases and document management systems for knowledge work process initiatives. The publisher created a database of text and figures. The high-tech manufacturer working on the documentation process was creating a documentation database. A computer software firm created a database of common human resource management questions. The airline created a database of complaint resolution letters and a work-flow management system for tracking complaints and responses. Still, none of these systems are conventional databases containing highly structured data fields. Knowledge-oriented systems are clearly more likely to involve text and less structured information. As a result, the most promising technologies for knowledge management are tools such as Lotus Notes or the World Wide Web, which our research sites were just beginning to develop.

Many firms wanted to use technology to extract knowledge from their vast quantities of information, but some failed to account for user requirements and behavior. For example, one engineering manager hoped that a new database for notifying product developers of approved designs and changes would influence product development activities down the road. He wanted to provide engineers with detailed product histories on previous designs. Yet there was little understanding of how the knowledge workers would utilize the information. Individuals who managed the release of the changes, not the product engineers who were in control of the product development process, had designed the engineering release system. As a result, the system was rarely used.

Implications of Different Redesign Strategies

As with improvement approaches, a firm might pursue one redesign strategy over another based on the orientation of knowledge work. For example, the application or reuse of knowledge can sometimes be enhanced by applying technology. However, when the improvement goal is better creation of knowledge, design strategies involving where and with whom people work are more likely to be effective. (Table 4 identifies the relationships between knowledge orientation and primary design strategies.)

Knowledge orientation is only one aspect in the choice of a design strategy. Other factors such as culture, strategy, competitive environment, and IT infrastructure can significantly affect design efficacy. By establishing the relationship between knowledge work process objectives and organizational characteristics, a manager can begin to sort out which designs to pursue and which to deemphasize.


Firms must better manage two of their most precious assets: knowledge and the people who create and possess it. Firms attempting to make their knowledge work processes more efficient and effective face a choice. They can adopt reengineering methods and approaches that have been employed for administrative and operational work. Or they can employ more traditional approaches that rely on knowledge workers to design and evaluate their own activities. In most cases, however, we believe that organizations will benefit by choosing an intermediate participative course between the two extremes. Using the strategies we have discussed, companies can select methods and tactics that reflect the type of knowledge work they are addressing, their organizational culture, and the business requirements for the change project. Of course, improvements to knowledge work are only one effort in a broad portfolio of improvement and change initiatives that managers must integrate.32


1. P.F. Drucker, The Age of Discontinuity (New York: Harper & Row, 1978).

2. U.S. Department of Labor Report, 1991.

3. “State of Reengineering Report” (Cambridge, Massachusetts: Computer Science Corporation, 1994).

4. T. Sakaiya, The Knowledge-Value Revolution (New York: Kodansha International, 1992).

5. J.B. Quinn, P. Anderson, and S. Finkelstein, “Managing Professional Intellect: Making the Most of the Best,” Harvard Business Review, volume 74, March–April 1996, pp. 71–80.

6. B. Levitt and J.G. March, “Organizational Learning,” Annual Review of Sociology, volume 14, 1988, pp. 319–340.

7. C. Nass, “Knowledge and Skills: Which Do Administrators Learn from Experience?,” Organization Science, volume 5, February 1994, pp. 38–50.

8. I. Nonaka, “The Dynamic Theory of Organizational Knowledge Creation,” Organization Science, volume 5, February 1994, pp. 14–37.

9. G. Davis et al., “Conceptual Model for Research on Knowledge Work” (Minneapolis: University of Minnesota, MISRC working paper, MISRC-WP-91-10, 1991).

10. Mintzberg’s analysis of managerial work, characterizing it as highly variable with a start and stop nature, is consistent with our definition of knowledge work. See:

H. Mintzberg, Mintzberg on Management (New York: Free Press, 1989).

11. D. Leonard-Barton, Wellsprings of Knowledge (Boston: Harvard Business School Press, 1995).

12. T.H. Davenport, Process Innovation (Boston: Harvard Business School Press, 1993).

13. J.B. Quinn and M.N. Baily, “Information Technology: The Key to Service Performance,” Brookings Review, volume 12, Summer 1994, pp. 36–41.

14. For a broad discussion of the organizational implications of professional organizations, see:

H. Mintzberg, The Structuring of Organizations (Englewood Cliffs, New Jersey: Prentice-Hall, 1979), pp. 348–379.

15. See, for example:

J.A. Raelin, The Clash of Culture: Managers Managing Professionals(Boston: Harvard Business School Press, 1989).

16. Earl describes knowledge processes in the following way: “Knowledge processes are not necessarily executed routinely, but require some decision rules and assumptions about the environment and occasionally are referred to in a mystical way as processes. Examples include strategy-making and innovation. The latter are difficult to analyze, model, and predict. Their flows are irregular, interrupted, and sometimes chaotic. These characteristics can be both bad and good. They are, as implied, more intensive of knowledge.” See:

M.J. Earl, “The New and the Old of Business Process Redesign,” Journal of Strategic Information Systems, volume 3, 1994, pp. 5–22.

17. This information is from interviews with Chrysler executives and J. Bennet, “The Designer Who Saved Chrysler,” New York Times, 30 January 1994), pp. 1, 6.

18. For example, one author wrote, “Process objectives must be quantified as specific targets for change.” Another advocate of more traditional operational business change wrote, “It is essential that the breakthrough goal be bottom-line and measurable.” See:

Davenport (1993), p. 128.

See also:

R.H. Schaffer, The Breakthrough Strategy (New York: Harper Business, 1988), p. 70; and

S.L. Jarvenpaa and D.B. Stoddard, “Reengineering Design Is Radical; Reengineering Change Is Not” (Babson Park, Massachusetts: Babson College, working paper, 1995).

19. Computer Science Corporation (1994).

20. T.H. Davenport, “Business Process Reengineering: Where It’s Been, Where It’s Going,” in W. Kettinger and V. Grover, eds., Business Process Change (Harrisburg, Pennsylvania: Idea Group Publishing, 1995); and

M. Hammer and S. Stanton, The Reengineering Revolution: A Handbook(New York: Harper Business, 1995).

21. Quinn et al. (1996).

22. M. Hammer and J.A. Champy, Reengineering the Corporation (New York: Harper Business, 1993).

23. Davis et al. (1991).

24. Discussion moderated by Joseph Podolsky in Hewlett-Packard’s Information Systems Group, Palo Alto, California, 1995.

25. Hammer and Champy (1993), p. 207.

26. D.G. Ancona and D.E. Caldwell, “Cross-Functional Teams: Blessing or Curse for New Product Development?,” in T.A. Kochan and M. Useem, eds., Transforming Organizations (New York: Oxford University Press, 1992), pp. 154–166; and

D.B. Stoddard, “Information Technology and Design/Manufacturing Integration” (Boston: Harvard Business School, doctoral dissertation, 1991).

27. Most of the projects involved attempt to change microlevel activities, for example, and many also involved substantial use of outsiders in the process design. There was a general desire to evaluate teams or process groups as well as individuals, and to evaluate performance more frequently. However, most of these reengineering-oriented projects had not yet been fully implemented. It may be that over time these projects will face the common phenomenon in reengineering of “revolutionary design, evolutionary implementation” that has been identified in other research. See:

D.B. Stoddard and S.L. Jarvenpaa, “Business Process Redesign: Tactics for Managing Radical Change,” Journal of Management Information Systems, volume 12, Summer 1995, pp. 81–107.

28. Of course, in business processes other than knowledge work there is also fragmentation of expertise. However, the firms we have observed in reengineering initiatives do not feel as compelled to represent all variations on the design team.

29. M. Burawoy, “The Anthropology of Industrial Work,” Annual Review of Anthropology, 1979, pp. 231–266; see also:

J.C. Lave and E. Wegner, “Situated Learning: Legitimate Peripheral Participation” (Palo Alto, California: Institute for Research on Learning, 1990).

30. The firms were a subset of the thirty firms that participated in our research project on knowledge work improvement (see Table 2).

31. I. Nonaka, “The Knowledge-Creating Company,” Harvard Business Review, volume 69, November–December 1991, pp. 96–104.

32. T.H. Davenport, “Need Radical Innovation and Continuous Improvement? Integrate Process Reengineering and TQM,” Planning Review, volume 21, May–June 1993, pp. 6–12.


This research project was supported by the Ernst & Young Center for Business Innovation and by the sponsors of the “Mastering the Information Environment” research program. We are grateful for their financial assistance and for their willingness to serve as research sites.