posted 1 Nov 2002 in Volume 6 Issue 3
Making knowledge work productive and effective
The nature of knowledge work is an area often ignored by firms looking to implement a knowledge management programme, yet real gains can be made by focusing on particular types of knowledge workers and targeting interventions accordingly. Thomas H. Davenport describes the experiences of Partners Health Care System to illustrate how knowledge work can be made more effective.
More than 30 years ago, Peter Drucker declared, “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.” We have yet to achieve his goal, but there are still good reasons for us to carry on trying. Knowledge workers are difficult to define and to count, but they are undoubtedly a major component of – and perhaps a majority in – US, western-European and other advanced economies.
I would define knowledge workers as those with a high degree of education or expertise, whose work primarily involves the creation, distribution or application of knowledge. Some knowledge workers have high levels of autonomy and discretion in how they do their work; others have more structured roles. Their activity – which includes R&D, marketing, engineering, planning, customer service and management – is critical to innovation and growth. Previous productivity-oriented programmes, such as process re-engineering and quality, didn’t really address knowledge workers; they focused on operational and administrative work. Even when some lower-level knowledge work has been ‘automated’ – as is the case, for example, with many call-centre workers – it often lacks quality and the ability to satisfy customers. So how can we enhance our economy by simultaneously increasing the speed and quality of our most valuable labour force?
Changing where and with whom people work
First, it’s critical to point out what not to do. In a 1995 research project originally focused on re-engineering knowledge-work processes, it became abundantly clear that the ‘r’ word didn’t apply. Managers of R&D and marketing departments acknowledged the need to improve processes, but cringed at the prospect of top-down, radical change. Participation in designing work is probably a good idea for all types of workers, but it seemed particularly important for autonomous, independent knowledge workers. My fellow researchers and I also realised that breaking down processes into detailed task flows wouldn’t work either. The knowledge workers we interviewed often felt that their work was too unpredictable, too unstructured and too variable to be diagrammed on a flowchart. In short, few of the traditional process-innovation or improvement approaches seemed to apply.
The approaches that did seem to work in the 1995 research appeared out of phase with the overall direction of business life – that is, with virtual and global organisations. When we asked successful managers of knowledge-work processes what they’d done to improve their results, they mentioned rudimentary techniques like putting all workers who needed to design a car into the same building, and in the same part of the building (for instance, at DaimlerChrysler). Toyota spoke of using a ‘big open room’ to get all the functions needed to build a car talking with each other. Mobil (now Exxon Mobil) described a team-based parallel approach to bidding on oil exploration leases, rather than an individual, sequential approach. My colleagues and I concluded that one relatively simple approach to improving knowledge work involved changing where and with whom people work.
In the 1995 research, we found that technologies of various types were also being investigated for their potential to improve knowledge-work processes. Most managers mentioned relatively simple technological tools, such as e-mail and databases, as being productivity aids. These were rarely combined, however, into a structured technology environment for the knowledge worker. This research was undertaken after the period of excitement about so-called ‘expert systems’, and just as the enthusiasm about knowledge management began to build.
In 1995, companies were beginning to implement knowledge repositories under the banner of knowledge management. They were interested in capturing the knowledge used in work, in sharing knowledge around the organisation and injecting knowledge into key processes. Many companies have since built these repositories, but my overall assessment is that they have not been very successful from the standpoint of transforming knowledge work. Particularly in today’s difficult economy, most knowledge workers just don’t have the time to access and sort through repositories that often contain large quantities of documents with variable quality. Technology still has an important role to play in knowledge work, but it needs to be integrated with and embedded into day-to-day activities. One key success factor, in addition to technology, lies in understanding the different varieties of knowledge work and applying different interventions to suit each type.
Segmenting knowledge work
What has become clear through more recent research is that it is not an effective approach to try to improve the productivity of all knowledge work in the same way within an organization. An intervention to improve productivity that might work for a customer-service representative in a call centre is unlikely to fit with the daily tasks of a research scientist. Furthermore, not all types of knowledge-work roles are of equal importance to organisations. Therefore, managers wishing to enhance productivity need to segment their knowledge workers. They need to identify the different types of knowledge work that take place in their organisation, in addition to the kinds of interventions that are likely to make sense. They also need to identify the knowledge-work jobs that are truly mission-critical – those roles in which increased productivity is critical to achieving strategic goals.
One such segmentation scheme is shown in figure 1. In this model, knowledge work is segmented into four types by the level of collaboration in the job, and the degree of judgement and expertise needed to perform the work successfully. Some diversified firms might have all four types, or even more based on other dimensions. Interventions – particularly those involving information technology, but perhaps other types as well – should be based on which of the four types of knowledge work are involved.
Figure 1 – a segmentation scheme for knowledge work
The key to improving productivity is to tailor the intervention to the type of work, and to the information and knowledge-related objectives for improving that class of work. For transaction workers such as those in call centres, the goal is typically to improve throughput by structuring the flow of work and ensuring that all of the information and knowledge necessary to do the job is easily available. All work is typically mediated through a computer, and there are clear, straightforward measures of productivity, such as calls per hour. In such settings, it’s common to ‘script’ work using databases and workflow automation tools. A call centre representative is often told what to say under any given circumstance, and access to knowledge resources is tightly constrained by the need for heads-down productivity.
This type of solution would not apply well to integrated workers, however, for instance engineers, programmers, and workers on large projects of multiple types. They typically have a substantial education and view themselves as professionals. Therefore, being told what to do at every turn by a computer would probably not go down very well.
One key approach to improving productivity among integrated knowledge workers is to encourage the re-use of existing intellectual assets, such as engineering designs for components, or pre-existing programming code for computer programmers. Yet these workers often enjoy creating new assets – and are rewarded for doing so – more than re-using existing ones. Therefore, integrated workers need templates and job aids that make it easy to re-use assets, and that gently guide them in the direction of re-use. Major automobile companies such as GM and Ford have found that engineers can’t be forced to use these assets, but will do so if it’s the path of least resistance. Another approach to productivity for this class of knowledge workers is to improve collaboration and to attempt to structure more integration across boundaries. At this point in time, though, the appropriate technologies and management approaches to achieve this objective have yet to be introduced.
Collaborative workers each have a high degree of expertise, but must co-ordinate their work with other knowledge workers. This knowledge-work category would include, for example, investment bankers, attorneys who work on large cases and consultants who work on large projects. Their job processes involve too much uncertainty and variation to be scripted or even fully mediated by computers. At this point, collaborative workers are best served by a series of tools that they can draw upon as needed; the individual worker remains the integrator of the tools and knowledge. While some firms have attempted to develop a type of ‘consultant’s workstation’ or automated ‘investment banker’s assistant’, these tools have not been broadly successful.
Interventions for expert workers may be the most difficult to attempt, because such workers have historically had a great deal of autonomy, and their jobs have had little structure. Experts do not generally have their jobs mediated by computers. Such workers also care a great deal about the quality of their work, and improved quality must be an attribute of any improvement effort. Experts often need to draw upon a high degree of knowledge in their work, but have historically relied upon their own education and offline knowledge resources (eg, articles and books) for the knowledge they require.
An expert intervention at Partners Health Care
It is possible to find successful IT-enabled interventions in the expert category, however. Partners Health Care System in Boston, which consists of several Harvard-affiliated teaching hospitals, implemented an intelligent physician order-entry system that adds intelligence to the patient care process while preserving the ability of physicians to overrule the system’s recommendations for medications, tests or referrals. I will describe this approach in detail, because if the work of these world-class physicians can be made significantly more productive and effective, other types of knowledge work might be transformed as well.
There are a variety of ways to bring knowledge to physicians in the course of their work, and Partners employs several of them. At the heart of its approach, however, is a computerised physician order-entry system, with trusted knowledge built in. When a doctor prescribes a drug, orders a test, refers a patient to another physician or even calls up a patient’s medical record, logic modules and a knowledge base are invoked to intervene in the care process. The system may inform the physician that the patient is taking a drug that would interact with the drug being prescribed, or that the drug prescribed is not effective or economical for the indicated disease. In the case of test orders, it may note that the test is not generally useful in addressing the disease or symptoms identified, or that the test has already been performed on the patient sufficient times to indicate a diagnosis or treatment. The systems may suggest that referrals are incorrect or unnecessary. Calling up a medical record may lead to a recommendation that certain follow-up tests or recommendations are desirable.
The physician’s autonomy is preserved because they can decline to accept the computer’s recommendation. The physician may know of circumstances of which the computer is not aware, or may choose to accept potentially negative consequences because of a more important benefit from the intended order. Preserving autonomy is critical not only to ensure that experts will accept the system, but also because the best results in knowledge work are usually obtained by humans and computers working together.
The order-entry system is key to the process because it sits at that point where knowledge comes into play. Without it, there would be no easy way to apply knowledge at the time and place where it is needed. Such order-entry systems may also be useful for efficiency, and for added safety in avoiding misinterpretation of poorly written orders. But the primary value is surely the ability to insert knowledge into the process at the crucial point.
There are, of course, times when patients need medical knowledge when there is no order being given. For these circumstances, Partners has developed a patient event-detection system, which provides alerts to physicians when a hospitalised patient’s monitored health indicators depart significantly from those expected. The physician is alerted of the changes to the patient’s condition through a wireless pager, and can then proceed to directly observe the patient or call in a new treatment. The order-entry system can also generate reminders to physicians that a particular patient should receive a call or schedule an appointment for a follow-up.
The power of both the order-entry-based knowledge and the event-detection system is that they operate in real time. Knowledge is applied directly and immediately to the patient-care process without the physician having to seek it out. In some situations, physicians can also get real-time access to experts for an online or telemedicine-based consultation.
Partners has also assembled many other sources of knowledge that are not real-time. This knowledge can be more extensive that that found in the logic modules, but it requires some time and motivation to seek it out. Online knowledge repositories include online journals and databases, care protocols or guidelines for particular diseases, the interpretive digests prepared by Partners physicians, formularies of approved drugs and details on their use, and even online textbooks. All of these knowledge resources are accessible through an integrated intranet portal.
While the resources available to Partners physicians are perhaps more extensive than to practitioners at other hospitals, similar resources are widely available. Most Partners physicians also do research, so they may have a relatively high appetite for online knowledge. They are busy people, however, and their time to consult such resources is certainly limited. At one Partners hospital, for example, the online knowledge repositories are consulted about a thousand times a day; at the same hospital, however, there are 13,000 orders per day that can benefit from the order entry-based knowledge.
While the Partners knowledge approach has been under development for over a decade, it is still not complete. The online order entry system and related knowledge are only accessible within the organisation’s two flagship hospitals, Mass General and Brigham and Women’s, but not for other hospitals or outpatient care centres. Medical knowledge for all of the diseases Partners physicians treat is yet to be codified. There is still plenty of work to be done.
Yet there is clear evidence that the approach the organisation has adopted is beneficial. A controlled study of the system’s impact on medication errors found that serious errors were reduced by 55 per cent. The frequency of prescription of a new drug that Partners experts found particularly beneficial for heart problems increased from 12 to 81 per cent. Another drug’s prescription frequency increased from 6 to 75 per cent when the system began recommending it because patients are required to take it fewer times per day. When the system began to remind physicians that patients prescribed a treatment of bed-rest also needed a prescription of the blood thinner heparin, the frequency of prescription increased from 24 to 54 per cent. These improvements not only save lives, they also save money. Recommendations from the system can point out drugs, for example, that are cheaper in addition to being more effective. Each adverse drug event costs over $2,000 to treat, so the system has clearly reduced the overall cost of treating patients.
Despite the virtues of the Partners system, it would not be appropriate for all types of knowledge workers – not even for nurses or administrators in the same hospitals, for example. The key to success is to design an intervention that is appropriate to the particular needs of a specific knowledge-worker role. This requires that organisations identify the knowledge workers within their organisations whose productivity and quality is critical to the organisation’s success, because of the expense and difficulty of transforming knowledge work. At Partners that was the job of the physician; for other organisations it is likely to be a very different role.
For too long, knowledge work has been ignored by organisations seeking to enhance productivity. Yet this area is too large and too important to the success of today’s firms to bypass. By segmenting and targeting particular types of knowledge workers, and designing technology applications that embed knowledge into day-to-day work, we can begin to achieve a revolution in post-industrial work.
1. Drucker, P., The Age of Discontinuity (Harper & Row, 1969)
2. Davenport, T.H., Jarvenpaa, S. & Beers, M., ‘Improving knowledge work processes’ in Sloan Management Review (Summer 1996)
3. Davenport, T.H., Thomas, R.J. & Cantrell, S., ‘The mysterious art and science of knowledge worker productivity’ in Sloan Management Review (Fall 2002)
Thomas H. Davenport is director of the Accenture Institute for Strategic Change. He can be contacted at firstname.lastname@example.org