In the modern enterprise, the business owner is usually concerned with five specific relationships:

The relative weight of these relationships changed dramatically in the second half of the twentieth century. In the past, only certain departments dealt with customers and vendors. The dominant relationship for most people was with the boss, with the relationship to subordinates coming in a close second. Immediate colleagues were part of the employee's world, but these were relatively few. Most communication was up and down the chain of command. Education about one's specialty and one's industry was via books and university courses outside the company.

This picture has now changed. Thanks to the wonders of modern communication, each employee is in direct contact not only with many more fellow employees, but also with colleagues outside the company who are specialists in the same fields or participants in the same industry. Moreover, decisions do not always have to be made by going up and down the organization chart. Decisions instead are now often based on collaborations among people from what used to be disparate organizational units.

Now many specialties are pursued by organic groups of people from many enterprises all over the world, who, on a daily basis, exchange ideas electronically about the best ways to do things.

Most significantly, nearly everyone has some direct involvement with clients, customers, and vendors—or at least has the facility to do so. Probably the most important relationship of all has become that of every employee to the company's customers.

To put this cell of the Architecture Framework (the business view of people and organizations) into perspective, it is important to understand the remarkable place at which we stand in human history. That history is, of course, far more complex than can be described in a few pages here, but a few points are worth noting.

Before the nineteenth century, people worked primarily on farms or as single artisans producing products one at a time. Wealth was measured in acreage of land. This period, which lasted for millennia, is often called the agricultural age. Early in the nineteenth century, things changed dramatically. Now people began to work in factories, producing hundreds or thousands of copies of the same product at once. Instructions were passed down an organizational hierarchy, and performance monitoring was passed up the hierarchy. The nineteenth and the first half of the twentieth century have been called the industrial age.

The driver of this new economy changed from land to capital. Where before, wealth was determined by an aristocrat's land holdings, now anyone could produce wealth with a factory, if one could accumulate enough capital to build it.^[1] The people who worked in these factories simply carried out the owner's instructions for making wealth. Where before, serfs did the labor that created the wealth from the land, now factory workers played the same role, creating wealth from capital.

In response to all this, Karl Marx, Charles Dickens, and others wrote of people's alienation when they worked simply as appendages to machinery. People didn't own the equipment they used. They were interchangeable. The jobs were narrow and offered no intrinsic satisfaction. Divisions grew up between owners and management on the one hand, who wanted the most output for the least money, and labor on the other, who wanted at least a living wage and respect for their efforts.

In spite of these conflicts, because of the nature of the work to be done, this remained the most economically attractive alternative well into the twentieth century, at least in the west. The Soviet Union tried to change the premises in 1917, and while the factories were then owned by the State instead of by private capitalists, and some attempts were made to provide for the general welfare, the underlying dynamic was no different. The economies were different, but the economics didn't change, and the work was still alienating. Eighty-four years later, the experiment was abandoned.

In the second half of the twentieth century things changed again. Suddenly information and knowledge became more important than physical capital. A company that is smarter in getting the most use out of a physical device will be more successful than one that is not. Marx's observations are no longer relevant, at least in the developed world, because the relationships between labor and capital have fundamentally changed. We each own our own knowledge, and this knowledge turns out to be the company's most important asset.^[2]

For this reason, the age we are now in the information age. There are those who would call it the knowledge age.

Actually, as Jonas Ridderstråle and Kjell Nordström asserted, Marx, Ho Chi Minh, Lenin, and Chairman Mao Zedong were all right. “They were right because they subscribed to the Marxist view that the workers should own the major assets of society, the critical means of production. We now do. And, perhaps, we did all along but we just didn't have the insight to realize it” [Ridderstråle and Nordström, 2000, p. 17].

While factories certainly have not gone away, fewer people work there, and even working there has changed. Much less of the work is done with unskilled labor. More of the work in manufacturing, for example, requires technical (knowledge-related) skills.

Now the worker chooses what to work on and how to go about it. Because the company is dependent upon the worker's knowledge, it must permit this to happen. It is in the nature of knowledge that it is communal, so people are no longer working on isolated tasks. The working environment is becoming clusters of people who share an area of interest or an objective.

Most significantly, their motivation is now in the work itself, not just the benefits bestowed by the corporation.

We own the means of production because those means are in our own heads. The physical capital on the balance sheet is no longer as important as the collective knowledge of the enterprise. The management of knowledge itself has become the primary focus of progressive companies today.

Thomas Stewart, in his 1997 book by that name, lists three kinds of “intellectual capital”:

Each of these is critical to a modern company's operation, but none of them show up in its chart of accounts.

A company often has much more knowledge and expertise than it realizes. Many companies are very poor at exploiting the knowledge they have. Traditional corporate organization has often prevented companies from gaining full benefit of employees' expertise. Many companies hold thousands of patents that they don't even know they own.

Microsoft is one of the most successful companies of all time, yet it produces virtually no physical product and has relatively little physical capital. True, it does deliver physical media, such as compact disks, but customers are not buying the media. They are buying the knowledge that is encoded on the media. Consider the microchip in your computer: The value of all the chips produced in a year exceeds the value of all steel produced in the same period [Stewart, 1997, p. 13].^[3] The value is not in the material, of course: it is in the design of the chip and in the design of the complex machines that make it. The value is in the knowledge required to build them.

Even companies that sell physical products, such as automobiles, have had to radically increase the intellectual content of their products. To compete, a car must now be more sophisticated, economizing on weight, cleverly getting the engine not to emit harmful gases, and providing just the right “feel”. All these things come from the auto maker's investment in knowledge and expertise, not from its investment in steel and rubber.

In 1998, The Berkshire Hathaway Company's net worth was $57.4 billion, the largest of any American corporation. The company sells insurance and other financial instruments, plus shoes, jewelry, other manufactured products, and ice cream. The company's market value, however, was only one-third that of knowledge companies Microsoft and General Electric [Berkshire Hathaway, 1998, p. 4].

The second component of intellectual capital for an enterprise is the technical infrastructure that makes a company's knowledge accessible. This includes everything from the Internet—for sharing ideas and thoughts on various subjects—to data warehouses that publish the operational data for a company. Success in accumulating structural capital directly affects an enterprise's success in the marketplace. Wal-Mart, for example, has revolutionized the retailing business by building what may be the world's most sophisticated information systems to support it. The company became the nation's largest corporation (in sales) in 2002, while during the same year its nearest competitor, K-Mart, declared bankruptcy.

In the days of smokestack capitalism, the economy consisted of factories producing thousands of copies of the same thing. Marketing consisted of persuading lots of people that the thing was exactly what they wanted. The customer was at the mercy of the producer.

Now, the balance of power is devolving onto the customer. Customer relations is turning out to be the “hot button” of modern business. Customers expect custom-made products. (Land's End just published an ad for swimsuits that are cut precisely for your shape.) This means that the company's relationship to the customer—its ability to understand clearly what the customer wants—is critical. Companies that have established such relationships are worth a great deal more than companies that have not.

So, what does all this mean to those of us who build systems? Knowledge management can be divided into two areas. Natural knowledge management is concerned with the way people learn and communicate with each other directly. In the past it has not been concerned with technology, but this is slowly changing. Artificial knowledge management is all about information processing using technological tools. This is the current concern of our operational systems and our data warehouses.

While we in the information industry have traditionally focused on artificial knowledge management, natural knowledge management is important too, and our success from now on will be measured in terms of our ability to manage both.

As we provide tools to support knowledge management, we must keep in mind these things:

The study of knowledge is both very old and very new. Philosophers have been writing about it for millennia. But attention to the relationship of knowledge to the structure of the workplace is relatively new. Most of what has been written about this topic has been published since 1990.

The changes described in this chapter have had a profound effect on the nature of the workplace. The traditional model of an enterprise has always been its organization chart. This has sometimes been a reflection of the structure of the work that the enterprise does, but not always. The boss wanted something and transmitted instructions to the employees. Their messages were then returned to the boss, describing how successful they had been in carrying out the instructions. The definition of work ultimately came from the boss. The worker contributed very little to the definition of the task.

For factory workers carrying out well-defined tasks this worked reasonably well. The boss described the tasks and the employees carried them out.^[4]

Where the tasks are intellectual, however, the hierarchical approach doesn't work at all. As “knowledge workers”, many of us no longer work for a “boss” who simply tells us what to do and makes sure that we do it to a specification. The boss may describe an objective, but the employee is now responsible for understanding the objective, plus the business environment, available technology, constraints, and the impact of various options on achieving that objective. The boss cannot possibly know all that is involved and is dependent on the worker's knowledge and skill.

Many of us have become “consultants” (even within our own companies), hired to assist “clients”, using our expertise and knowledge. The worker is now in the position not only of assisting the client in achieving objectives, but also of telling the client whether the objective is reasonable in the first place and, if it is not, recommending alternative objectives.

Instead of focusing on the politics of keeping various bosses happy, our entire transaction with the client now comes down to whether or not we are providing a useful service. Not only is the client free to let us go if we are not being useful, but we probably want to go if the environment is not one where we can be productive. This is a much happier arrangement than we have known in the corporate world, where we always had to be alert to making the right people happy, in ways that often had nothing to do with our skills or abilities—or the job at hand, for that matter.

Our motivation is no longer security and money.^[5] We work on projects because they stimulate our imagination and intellect. We will work for a company as long as it provides interesting and meaningful projects. When it stops doing so, we will go somewhere else.

This has a profound effect on how companies are managed. Suddenly managers must be alert to the dynamics of their departments in ways they never were before. This has not been easy. For years, employees at a Canadian software company lobbied management to revamp their products to make them more integrated, more reliable, more robust, and more effective. Their pleas fell on deaf ears, but at one point the company published a brochure, falsely claiming that the products had exactly the qualities the employees were trying to promote. Management's intransigence, along with its lack of understanding of what the employees were trying to do, caused 20 of the company's 25 employees to quit. Twelve of them created a competing company.

On a happier note, one company has launched a four-year plan and has organized its personnel entirely around it. It is a matrix, where the rows are groups of projects and the columns are areas of expertise. There are three columns that are the basis for maintaining quality work: Applications, Data, and Technology. The rows address such project areas as “People and Organizations”, “Physical Assets”, “Finance”, and so forth. The people, then, all work at the intersection of a row and a column. The idea is that each person is supposed to be working toward implementation of one or more projects as expeditiously as possible. But each also has the obligation to do quality work, as defined by the column involved. Data people are obligated, for example, to follow the best practices in industry in organizing and acquiring quality data—even if it may affect the project plan.

This is one example. Other companies have different kinds of matrix organizations (or hierarchical ones) as well.

Thomas Stewart describes the opinion of Frank Walker, president of GTW Corp., that in the years to come there will ultimately be only four types of career:

The fact of the matter is that the business owner's view is of a world that is changing under our feet at an astounding rate. This affects our personal lives, how we relate to our work, and how our employers relate to us. We have incredible new opportunities for creativity and self-discovery, but it comes at the cost of an overwhelming amount of uncertainty and chaos. Companies must deal with this, and we must deal with it as individuals. Is there anything out there that can help us come to terms with all this change?

As it happens, there may be....

To properly address the problem of variety, it is necessary to discuss systems. We are not speaking now of a system as a collection of computer components, but rather as any arrangement of components that is self-perpetuating. This includes everything from corporations and governments to the human body. (Mr. Beer's original work was in physiology, and many of the insights he presents came from that field.) The enterprise, as described by John Zachman, for example, is a very good example of such a system. Mr. Beer's work is concerned with the quest to understand the fundamental structures and phenomena that govern any such system.

One filter used to reduce variety in order to manage a system is to divide it into smaller parts that have less variety. The question is, how to do that? It turns out that the only meaningful approach is to recognize that all systems are fundamentally recursive. That is, each system is a complete system within another system. In turn, each system contains one or more other systems. As you look at the world around you, you will discover that this is the way it is organized. Your family is in a community; the community is in a state; and so on. Every system you can name is part of and managed by a larger system.

Since variety depends at least partly on the purpose of the system exhibiting it, what is the purpose of any particular system? Because systems are so various, from British Rail to an English Sheepdog, it is reasonable to ask, what is the purpose of each? Upon close analysis it turns out that they all have the same purpose: to continue to exist. Virtually all of the actions taken by any system are directed to that end.

As defined above, cybernetics is the science of communication and control. Mr. Beer makes extensive use of the works of H. Ross Ashby, a cyberneticist who wrote in the 1950s. Ashby postulated a series of natural laws that apply to any control mechanism. He proposed that control mechanisms must follow these laws if they are to work, just as airplanes must account for the laws of aerodynamics if they are to fly. An excellent example of his work is An Introduction to Cybernetics, written in 1956, which explains these laws in very clear language.

Mr. Beer takes Dr. Ashby's work a step further when he then asks the question:

We can use the laws of aerodynamics to determine characteristics required of an airplane to allow it to fly. How can we use the laws of cybernetics to determine the characteristics required of a system to allow it to continue to exist? Expressed in another way, what are the characteristics of what he calls a viable system? Specifically, how can variety be managed so that the system can get on with its work?

What does this have to do with people, by the way? As it happens, in any organization, it is the people that are the organization's fundamental components. As Mr. Beer puts it, people are The Heart of Enterprise (hence, the title of his work). At any level, a person's job in an organization is to manage variety.

The assignment to understand what makes an enterprise viable is an assignment to understand the required nature of interactions among human beings, and how they collectively manage variety. How can a system be divided so that the variety of each component is manageable?

How do you manage a system? The basic cybernetic model, shown in Figure 5.2, has a process of some sort,^[7] with a controller—an agent controlling it. It also has information channels for communicating between the controller and the process. The controller makes use of a set point, a reference value that serves as a target value for the process. The process generates information (values of specified variables) about its performance which is communicated through a channel to the controller. This information is continuously compared to the set point, and, depending on the result of the comparison, the controller may then send commands to the process through a different channel. If the values are too low, one set of commands is sent. If they are too high, a different set is sent.

Figure 5.2. The Feedback Loop.

The signals from the process to the controller constitute feedback. This configuration of process and controller is called a feedback loop.

Loops like this are most visible in physical control mechanisms like thermostats and manufacturing process control. In a thermostat the set point is a desired temperature. If the actual temperature goes below that, a signal goes to the furnace to turn it on. When it goes above the setpoint, a signal goes to the furnace to turn it off. In process control, a set point might be a value of a measured variable, such as oil pressure. If the oil pressure in the process goes too low (or too high, depending on the situation), a signal is sent to a motor to open (or close) a valve.

Feedback loops are also present in any organization. The manager has some vision of what is desired, and he issues instructions to the operational groups. Information about the organization's performance is then returned and evaluated. A sales target could serve as a set point, for example, such that if sales went below that target, steps would be taken to increase sales.

In manufacturing, this was formalized during the 1970s and 1980s with the integration of Master Production Scheduling, Materials Requirements and Capacity Planning, and Shop Floor Control systems. The master schedule laid out the products to be produced over time, Material Requirements Planning and Capacity Planning determined what materials and other resources would be required to satisfy the schedule, and Shop Floor Control provided feedback about what was actually happening so that the schedule could be adjusted as necessary.

In his search for the viable system, Mr. Beer expands on this simple model. He starts by pointing out that the system does not operate in a vacuum. Each process exists within its environment. The model then, actually has three components: the environment, the process (which he calls “operations”), and the managing controller. Because of some things he's going to do with the model later, Mr. Beer chooses to rotate it 90 degrees, as shown in Figure 5.3.

Figure 5.3. The Feedback Loop, Expanded.

Note that in this diagram management does not interact directly with the environment. Only the operational element does that. There is in fact an interaction between management and the environment, but it is of a different kind than is shown here and will be discussed further below.

At this point we introduce one of Mr. Ashby's natural laws of cybernetics: the Law of Requisite Variety [Ashby, 1956, p. 206]. That is:

A particular management has at its disposal only a finite amount of variety—a specified number of available actions to take. The Law of Requisite Variety means that management can only deal with an amount of variety equal to the amount at its disposal.

For example, if a process can assume 15 different states, the Law of Requisite Variety says that the controller must have a variety of exactly 15. This is reasonable, since, if the process assumes the 15^th state and the controller's variety is only 14, the controller won't know what to do. By the same token, if the controller is capable of a 16^th response, this is irrelevant.^[8]

If the variety in the world is increasing at the rate we perceive it to be, this puts a tremendous pressure on management to increase the variety at its disposal as much as possible. This can be done via education, management techniques, and other methods. Ultimately, however, management cannot keep up. This is also part of the natural law.

In Figure 5.3, the “V” in each component refers to the relative variety handled by that component. The world at large has a great deal more variety than your typical process operation, which in turn has more variety than management. Consequently, Environment's V is larger than Operations' V, which is larger than Management's V. Think of all the potential customers out there, with myriad tastes, which may well change from day to day. In your company, you have a finite number of products and options to offer for sale. No matter how sophisticated you become, you will never be able to match the variety of the world.

The operation, in turn, has more variety than management. Your typical manager is responsible for a department of perhaps 15 people, each of whom has a collection of ambitions, talents, and personality quirks. This is in addition to whatever technology is being used, the particulars of the department's operations, and much more. On top of that, there is the obligation to learn all the new things that are happening in the company's industry. There is no way that the manager has the variety needed to match the variety of the operation. The manager cannot respond to every event that takes place in the organization. The implications of sets of events must be understood.

The Law of Requisite Variety, then, requires the introduction of amplifiers and attenuators (filters) to make the variety perceived at each point the same as the variety at the disposal of each point. (See Figure 5.4.) In other words, an attenuator filters the variety to the exact amount that can be processed by the receiver. The business filters what it sees in the world. It perceives the market only in terms that it is capable of understanding. In Figure 5.3, this is the filter on the left side. Similarly, management filters what it sees of a company's operations. Ultimately the manager can perceive only as much as can be acted upon. In Figure 5.4, this is the filter on the right side.

Figure 5.4. Amplifiers and Attenuators.

Going in the other direction, an amplifier multiplies the effects of management's decisions to make them responsive to the variety of the operation (shown on the right side of Figure 5.4), just as another amplifier can multiply the effects of an operation to make it responsive to the environment (shown on the left side of Figure 5.4).

This recognition led Mr. Beer to formulate his First Principle of Organization:

To say that varieties equate is to say that the variety left after some is consumed by filters must be equal to the amount of variety that can be perceived by and dealt with by management. Similarly, the variety generated by management (which is equal to the amount perceived) must be augmented by amplifiers so that it is equal to the variety that can be dealt with by operations. As stated above, we are saying here that the incoming variety of a controller is exactly equal to the variety disposed by it.

The design point is at work when the filters are designed so that the controller sees the right set of things. A successful company analyzes the world and determines the two or three or four things it is capable of reacting to. It then designs the appropriate filters to provide information on just those things.

Similarly, the manager has to filter the information received from operations. If reports and display screens are well designed, the manager will be in a position to make intelligent decisions about the operation. If, instead, the more common filters “blind ignorance”, “prejudice”, “politics”, and being “intellectually challenged” are used, then the decisions may prove to be a bit less successful.

The inability to see things is itself a filter.

An amplifier takes the manager's simple decisions, such as “increase staff by 50%”, and expands their information content to incorporate all the specific implications of those decisions. Similarly, the enterprise's marketing efforts constitute amplifiers to disseminate the enterprise's message from operations to the world at large. The enterprise may advertise on TV, for example, to encourage many people to look at its website. The website, in turn, can communicate more than the advertising campaign could, further increasing the variety of the message.

To the extent that the company designs these filters and amplifiers well, it will be successful in the marketplace. To the extent that it doesn't, it won't.

If we are to design these amplifiers and filters, we must give thought to their characteristics. These are addressed in Mr. Beer's Second Principle of Operation:

The important element that this principle introduces is time. Is the channel sufficient to convey what must be conveyed in the time allowed? For example, there is a crisis in the plant, and among other things, the computer network is down. An indicator light in the main office communicates that “The Framis isn't working”. That's a filter all right, but it in fact filters out critical information required for the manager to respond. This is an example of a communications channel without sufficient capacity to transmit required information. More useful would be a telephone call saying “The Framis is down because of flooding!” That's a channel that can convey the information required.

As another example, consider a worker having trouble with an online system. She tries several things and nothing works. She is already considerably frustrated when she calls the help line. She tells them “This system is [expletive deleted]! I can't get it to work! I just can't!” This is a highly filtered message to the help desk, without enough information getting through to allow the help desk to determine the exact cause of the problem.

As another example, many top management reports are made up of averages of critical data, such as sales or expenses. But the averaging process may lose key points that are required to make an intelligent decision. In many cases, too, there are people lobbying and promoting particular policies, very carefully protecting the boss from meaningful knowledge about what is really going on.

These are examples of communication channels whose variety is not adequate to the task of managing the enterprise.

We can examine these amplifiers and attenuators in more detail through Mr. Beer's Third Principle of Operation:

That is, a translation (“transduction”) takes place when the operation communicates with the controller. The channel must not only have the requisite information capacity, but it must be able to deal with the specific states that are being transmitted, so that it can correctly translate the variety into terms management can understand. A Board of Directors listening to a senior manager's presentation might well have the information capacity to understand a very sophisticated presentation. But if the manager presents a gray, ambiguous situation in black-and-white terms, the translation process is being subverted.^[10]

We often make mistakes in our evaluation of people, thanks to the “halo” effect. Because people are well mannered and well dressed, we translate these signals into the assumption that they are therefore intelligent and knowledgeable. In fact we have perceived much less variety than is there for us to see, or than we could understand if our filters weren't subverting the translation process. Indeed, prejudice itself may be characterized as a problem of variety: we can see only two states—“us” and “them”.

OK, let's take this view of feedback loops and see if we can come up with a more sophisticated model of the viable system—one that will give us some guidance as to how to design those attenuators and amplifiers.

First, looking at the original configuration in Figure 5.2, what if something happens in the circular operation presented that is outside what the controller knows how to deal with? There is a fire in a plant, for example. This isn't in the procedure book. Once the mess is cleaned up, what is needed is a mechanism for reviewing the basic feedback loop and evaluating its set points and other structures to be sure that the loop can handle this situation next time. In this case, next time there should be a procedure in place for mitigating the effects of a fire.

An adjuster organizer is a mechanism for modifying a basic feedback loop's set points and other structures. (See Figure 5.5.) This enables the inner loop to handle a new situation. This adjuster organizer in the outer loop can adjust the feedback process itself, responding to things that the inner loop could not. In the fire example, the adjuster organizer implements the new procedure for mitigating the effects of fire. In principle, the adjuster organizer also is controlled by a set point, although this could be something as crude as “the company falls apart”. That would be a trigger for the adjuster organizer to act.

Figure 5.5. The Adjuster Organizer.

This brings us to the idea of recursion. Assume for a moment that we have a set of systems of the sort shown in Figure 5.4, and that they are inside some sort of meta-management that provides them with adjuster organizers to perform the necessary feedback adjustments, as shown in Figure 5.5. The adjuster organizer constitutes a meta-management. A meta-management is a controller at one level of recursion that is responsible for managing a set of systems at the next level of recursion down.

A set of systems, each with its environment, process, and controller, are represented in Figure 5.6 as what Mr. Beer calls System One. These systems exist in the context of some meta-management, whose exact nature we have yet to discover, but whose function will be to oversee the feedback loops of all the Systems One.

Figure 5.6. System One.

Figure 5.6 is a considerably simplified representation. The fact that there are only three Systems One is arbitrary, for purposes of this demonstration. Naturally, in a real system there could be more or fewer. In addition, the lines of communication from the first to the second and from the second to the third are not meant to imply that there are no lines from the first to the third. To draw them, however, would unnecessarily clutter the diagram. Similarly, the lines from the meta-management to each of the three Systems One are not meant to imply that the meta-management communicates only to the third System One via the first two. Again, in the interest of clarity, the detailed lines were left out.

In fact there are 18 channels connecting the various parts of this three-part set of Systems One.^[11] If there were four Systems One instead of three, the total number would increase to 34. “Thus does variety proliferate,” says Mr. Beer.

(In the diagram, for neatness' sake, the “1” appears only on the management box. In fact, each entire assembly of environment, operations, and management constitutes a System One.)

Note that the dashed lines showing connections between operational circles may be strong or weak. They are almost certainly there. Labor unions, for example, have always been very good at communicating across operational units. Twenty years after Mr. Beer wrote his book, the technology of electronic mail has greatly increased the extent of these communications channels. The meta-management should know this and take it into account.

This brings up an important issue that has been part of the discussion of human systems over the centuries: the emotive concept of freedom. Much has been written about it and much passion has been expended. It turns out that cybernetics has a few things to say on the subject.

If you look at the model for System One, you see that each of the Systems One appears to be and ought to be autonomous. It would seem desirable for each to be able to react to the environment exactly as it sees fit. If we are talking about Systems One that are our households, in the metasystem that is our local government, this certainly seems to be the case.

And yet in both the operational and the managerial columns, it is reasonable to expect that a meta-manager is going to exert some interventions, if only to ensure that the Systems One don't work at cross purposes. The meaning of “cross purposes”, of course, is defined by the meta-management. The Meta-management seeks a cohesiveness that is not necessarily perceived as possible or even desirable by the Systems One, but which is directly connected to the meta-management's mission.

Thus the meta-management will intervene in the operation and management of Systems One. This intervention comes in the form of communication on the vertical plane between it and the Systems One, at right angles to the horizontal management of each System One. In intervening, the metasystem will reduce the variety available to each System One. A corporate salary policy, for example, constrains each manager as to what can be offered to employees as compensation.

The nature and extent of the intervention will vary, depending on the organization and the circumstances of the intervention. The extent to which freedom is perceived to be constrained will also depend on the organization as well as its perceived purposes. A military organization, for example, places extensive constraints on the range of actions available to soldiers, which constrain their freedom considerably. These constraints are perceived to be essential to the survival of the unit, however, so (for the most part) the soldiers go along with it. A commune, on the other hand, whose purpose is to spread peace and love, provides many fewer constraints on its citizens—consistent with its purpose.

It is interesting that the nature of the intervention's effect on variety is usually misunderstood. This can be to the detriment of the organization. For example, a project that is intentionally not autocratically managed may require endless committee meetings to ensure that all interests are accounted for, and even these may be in the unwritten context that what is done must follow corporate policy, even if that policy isn't explicitly expressed. Participants wind up with less freedom than they would have if objectives were more clearly expressed and they could focus on those that pertained to each of them. A management intervention to clarify objectives in this case would increase freedom.

Mr. Beer lays out the following chain of argument as to how to deal with the issue of vertical interventions and their effects on the autonomy of Systems One [Beer, 1979, pp. 157–158]:

Therefore

The meta-management should make minimal use of variety amplifiers in its dealings with management units in a downward direction. But this is difficult, since even “policies” and “guidelines” tend to be perceived from the other direction as massive constraints.

This is the cybernetic argument for autonomy (or freedom), as distinct from the ethical, political, or psychological arguments. Unlike these others, it has no emotive content. It is basically mathematical.

Therefore

The meta-management must make some intervention, and should make only that degree of intervention that is required to maintain cohesiveness in a viable system. For a viable system cannot disintegrate without losing its viability.

Therefore

So, as with the horizontal management lines, the amplifiers and attenuators along the vertical axis must also be designed. If they are not, they will simply “grow”—not necessarily to the advantage of the organization. More significantly, they will grow in ways that are not perceived (and certainly not understood) by the managers involved.

In Chapter 8 is a model of an organization's ends and means. It specifically deals with what it calls “elements of guidance” (business policies and business rules) that amplify management decisions, and “assessments” that provide a filtered view of the effect of the environment (there described as “influences”) on the organization's ends and means.

In modern times there is a particularly pernicious “nutrient medium” in which this growth takes place—the computer. The computer allows us unlimited options in the way communications are established. When Mr. Beer wrote his book, the situation was bad enough, but in the 20 years since then, these options have increased exponentially. Our intelligence in employing these options, alas, has not increased nearly as much.

Thus, buried in the middle of Chapter 5 of this book is its most important point:

So, what does that vertical axis look like in the viable system?

We have established that it is desirable for each System One to have a certain amount of autonomy to conduct its own affairs, but it is also desirable to moderate autonomy and to promote their cooperation.

The first problem with a set of Systems One is their tendency to step on each other's toes or to oscillate. Oscillation occurs when the action of one System One causes an adverse effect on another System One. The reaction of the second exacerbates the situation in the first, which reacts in a way even worse for the second. This continues until the entire situation degenerates completely.

As an example of how subsystems can work at cross purposes, a production order is placed on production line 1 for widget A-6, which requires parts P-4, P-6, and P-8. It is due January 15. The parts needed are then allocated in inventory and therefore are not available to any other orders. An order for widget B-10 is placed on production line 2 for delivery January 17, requiring parts P-32, P-17, and also P-8. The production line 2 foreman discovers that (after the allocation to A-6) there are not enough P-8s to complete the order, although more are coming in on January 24. Therefore, the order for B-10 must be held for one week.

Meanwhile, however, it turns out the P-4 parts are defective. This means that A-6 cannot actually be made until more parts are received on January 30.

Since production line 1 and production line 2 don't communicate with each other, there is no way for production line 2 to learn that, in fact, the P-8 material is available and the order for B-10 can be completed.

This kind of problem is common in business, especially in manufacturing, and the informal communications between operations are often not sufficient to address it. What is needed is some oversight that can take into account the set of operations in the various Systems One and coordinate them. This argues for the creation of System Two. System Two's purpose is to control oscillation between pairs of Systems One.

In Figure 5.7 we have added triangles to represent this new (sub)system. Note that System Two does not instruct any System One as to what it should do, beyond what is required to coordinate between the different Systems One. For this reason, it is not part of the meta-management. This function is represented in the manufacturing world, for example, by materials management, whose job is to prevent precisely the disruptions described above. In this case, System Two would be aware of the requirements and limitations of both production lines and be able to release parts that could not otherwise be used. Without instructing the plants in any way as to how to manufacture, or even what to manufacture, it simply coordinates supply and demand to ensure that the right quantity is manufactured each week.

Figure 5.7. System Two.

Ah, you say, but, in addition to System Two, there still has to be some management over the set of Systems One to direct them to one or more corporate (metalevel) goals. Meta-management is still required to direct the set of Systems One.

In Figure 5.8 System Three appears as the first recognizable component of the meta-management. It is shown as a component of the meta-management rather than its entirety, because we don't know yet what other components might be present.

Figure 5.8. System Three.

Note that System Two is not part of the meta-management box, because it isn't really supervising System One. It is not at the meta-level of recursion. It is simply providing a service, and it lives at the same level of recursion as System One.

System Three's job is to integrate the operations of the collected Systems One. It deals with issues that each System One cannot see. Since it is desirable for Systems One to have as much autonomy as possible, this doesn't simply mean “you do what I tell you”. Rather, as described above, integration is the “minimal meta-systematic intervention that is consistent with cohesiveness within the purposes of the viable system. These purposes are not, of course, objective properties of the system; they are formulated by those who have charge of it, which include its initiators, its owners, its employees, its customers, and its observers” [Beer, 1979, p. 202].

As part of the meta-management, System Three is in communication with all the Systems One simultaneously. It knows what is going on inside the firm, now.

The actual interventions taken by System Three will depend on its purposes, but we can assume that one purpose will include getting the most out of the set of Systems One. That is, the objective is synergy, an increased output for the whole, even if it means directing that one or more of the components be less than fully productive.

It is important to pause here and remember what is being described. This model is not advising how an organization ought to be run. It is describing what must be present in any viable system. Normally systems (enterprises) aren't described this way. Mr. Beer has created a new language for describing how an enterprise must work. So far we have acknowledged that any system must be viable and that it must exist in the context of a higher-level system. Moreover, a set of viable systems must be regulated, first by a system that dampens oscillations and then by a system which generates synergy across the set of systems to achieve a meta-level goal.

Any viable enterprise must have these elements. One which does not will not long survive.

Note, however, that while these are necessary conditions for survival, they are not sufficient. What we have so far is an automobile that is idling. The component systems are operating and System Three is working to maximize their total output, but this in no way accounts for changes that will be necessary (at this level of recursion) to adjust to changes in the environment. What we have so far is not sufficient for the system to make “progress”.

What is required is innovation. It is necessary for a viable system to be constantly examining the environment to find new opportunities and to ward off risks.

In 2000, Jonas Ridderstråle and Kjell Nordström, two leather-clad Swedish college professors with shaved heads, wrote a very witty and entertaining book on the modern world called Funky Business. In it, they describe “Funky, Inc.”, their view of what the modern company should be. Among other things, “Funky, Inc. is extremely innovative. In a real-time, globally linked surplus society it is just a matter of a few weeks, days, or even hours before our friends from Bangalore, New York, Kuala Lumpur, Paris, Gdansk, Tokyo, Seoul, London, or Santiago come here to copy our recipes. To remain unique, we must constantly sharpen our competitive edge. Alan F. Shugart, Chairman of disk drive giant Seagate Technology Inc., goes as far to say that 'Sometimes I think we'll see the day when you introduce a product in the morning and announce its end of life at the end of the day' [Shugart, 1997]. And IT guru Kevin Kelly says 'Wealth flows directly from innovation... not optimization... wealth is not gained by perfecting the known' [Kelly, 1998]” [Ridderstråle and Nordström, 2000, p. 152].

To deal with this change, we must add a System Four. Figure 5.9 shows this. Here, System Four's primary communications are not with other parts of the system but with the environment. Note that there are really two domains in the environment that System Four must deal with. The first is the accepted environment, which is the part of the environment that Systems One already address. The second, more important part of the environment that System Four must be about is the problematic environment. It is in the second environment—the one not seen by Systems One—that System Four has the ability to identify the true opportunities (and risks) to the business. (Note the discussion of the “adjuster organizer” in Figure 5.5, above.) What things are going on in the world that the operating divisions are not in a position to see?

Figure 5.9. System Four.

In one real example, a book store conglomerate was initially unable to understand the implications of the World Wide Web. And yet that bit of environment has suddenly raised its ugly head and completely changed the business of selling books. It required a company with an active System Four to spot that. (More significantly, it took a brand-new company willing to look at the world clearly to see that. Existing companies, dominated by their Systems Three, could not for some time.)

If System Three is about “inside and now”, System Four is clearly about “outside and then”. System Three's domain is the system and its operations; System Four's domain is the world at large.

Note, by the way, that “innovation is not only a matter of technology—nuts and bolts. In fact, technology is often only a small part of it. Innovation concerns every little aspect of how an organization operates—administrative innovations, marketing innovations, financial innovations, HRM [human resources management] innovations, and service-concept innovations. Going for total innovation, therefore, requires rethinking every little aspect of how we operate” [Ridderstråle and Nordström, 2000, p. 153].

More significantly, innovation is not just a matter of listening to the environment and determining “what the public wants”. The truly innovative company provides goods and services the public didn't know it wanted. Did 3M ask the public if it should invent Post-it Notes? Did Starbucks ask the public if they wanted a cup of coffee available on every street corner? The trick to real innovation is to create new things and then make them widely available. Real innovation often involves educating the consumer.

All this talk about innovation raises one of the more interesting aspects of System Four. Even though we have emphasized here that the model does not correspond to an organization chart, it is not hard to find corporate analogies for Systems One and Three, as well as, with a bit more difficulty, System Two. Most companies, however, do not have a formal “System Four Department”. Some have “Research and Development”. Some have “Market Research”. Some have “Corporate Development”. For most companies, though, the real System Four consists of a random collection of insightful, clever, and forceful employees who could literally be working anywhere in the company. Somebody with enough nerve and aggressiveness may be willing to stir up trouble until someone in authority finally responds to a good idea. Unfortunately, the way most companies are organized, excellent ideas from people with less nerve and aggressiveness are often lost.

What we more often have is a market research department that has little contact with research and development. The information services department is disconnected from corporate planning. Each of these reports as a staff function to different parts of the enterprise. The existence of System Four as an essential condition for the enterprise's viability has often gone unrecognized.

An interesting side effect of the move toward knowledge management is that the System Four function is finally being dealt with in some companies. It has been hard to institutionalize it, but people are beginning to be appreciated and rewarded for their creativity and insight. Even so, in most real companies, this System Four functionality has not been adequately integrated. As Messrs. Ridderstråle and Nordström point out, this has profound implications on twenty-first century business—for two reasons.

The first reason is the incredible increase in the rate of change in the environment. When the world was relatively stable, a CEO could charge a small group of people to study a change in the world about them, and occasionally they would come up with a useful idea. This is no longer adequate. The survival of the firm depends on its ability repeatedly to respond and adapt quickly to its environment.

Not only is technology changing more rapidly than even the most agile enterprise can accommodate, but the very institutions and values of our newly global society are changing. This means that, among other things, markets are now international, consumers demand specific products for their needs, and the old structures of corporation and government are becoming obsolete. The company that cannot cope with these changes will soon no longer be in business.

For example, one company realized a few years ago that its status as a premier glass manufacturer in Mexico was insignificant if it was simply a minor player in North America.

The second reason for profound implications is that the allocation of responsibility for responsive behavior in the past is inappropriate for supporting true innovation. Before, innovation was mostly carried out by what were called “staff” functions—supporting management as a kind of extension of it. The idea of “staff” suggests the taking on of routine activities that managers are not interested in doing themselves. That does not describe what is required of System Four in the modern era. System Four is now concerned with technologies that are probably outside the sphere of competence of the boss. It must be carried out by people with more experience in specific areas than the boss, and with more autonomy than they have ever known before.

There is, of course, another basic problem with System Four: All these good ideas about the environment and the future often disrupt the nice, pretty equilibrium of System Three. This means that there will always be a fundamental conflict between Systems Three and Four. Implementation of integrated manufacturing planning techniques, for example, can make a company much more responsive to the marketplace and can significantly reduce inventory costs. Such techniques, however, can conflict directly with the corporate culture of manufacturing via more informal systems [Wight, 1975].

As a further example, it is ironic that in the 20 or so years since Mr. Beer wrote his book, a lot of the acceleration of technology has been in the information processing arena. The World Wide Web, for example, has completely changed the marketplace in which most companies operate. In principle, this means that much of a company's potential for innovation should reside in the information technology department. The problem, however, is that, while it may promote change in the information processing tools, the information technology department has often been the one where it is most difficult to absorb change. Systems in place are hard to modify, and the people working in information technology departments are often wedded to older structures. “We can't do that, because the systems we have won't permit it”. This will be the epitaph of many companies in the twenty-first century. How do we resolve this conflict?

Resolution of this conflict requires the introduction of the last component of meta-management: System Five. This system resolves differences in point of view between System Four and System Five, based on its understanding of the identity of the viable system.

System Five is the only one that rates its own Row in the Architecture Framework. It represents the identity of the firm. It defines the enterprise's scope. We are a bank and not an oil refinery. What does that mean? What does it mean about our data, functions, planning cycles, locations, business rules, and yes, our people? System Five is the part of the meta-management that knows why we are in business, and therefore System Five is the ultimate arbiter of disagreements between Systems Three and Four. (See Figure 5.10.)

Figure 5.10. System Five.

This doesn't mean that being System Five is easy, especially these days. The pressures of the global marketplace challenge everything about the way companies have been run. Responding adequately could mean wrenching changes that may be difficult or impossible to carry out. But it is System Five that must ultimately make the decision whether or not to proceed.

The inclusion of System Five completes the definition of meta-management. Now we can see how the recursion works. This meta-management is really the controller of a System One at recursion's next level. (See Figure 5.11.) Note that, contrary to what was said at the beginning of the chapter, in fact there is communication between the controller of System One and the environment, but as we saw above, it is of a different quality than the communications from the process and the environment.

Figure 5.11. The Next Recursion.

Notice that what are represented here are system functions, not necessarily an actual physical organization. For example, each operating division might have its own Finance Director, Marketing Vice President, and Manufacturing Management. The same positions might exist in the corporate management. It is not unreasonable for the corresponding functions to meet across divisions, perhaps with the corporate manager sitting in as well. In this case, it is the group that is serving as System Three.

In a small organization, a single person might serve as Systems One, Two, and Three, although it is important for all to realize that any particular message from that person to the operating groups is either a System One message, a System Two message, or a System Three message. In a one-man consulting organization, for example, the entrepreneur may decide what kind of business to pursue (System Four), schedule projects so that they don't overlap (System Two), and perform the work (System One).

One way that the organization chart is misleading is that each box on it does not necessarily represent a viable system in its own right. It may simply be a component of the parent, without having any of the characteristics of a viable system that are being promulgated here. The Systems One depicted in the model here do not correspond to such things as “finance”, “sales”, and “manufacturing”. Each System One must itself be a viable, self-contained system. This means that manufacturing and its connection to the outside world, sales, must be part of the same System One. Dissecting a System One into subsystems, based on the tasks each performs, is not the same as identifying its component Systems One.

As much as the independent consultant, for example, would love to have the “company” concerned only with using skills, it can't be done. The entrepreneur must also do all the tasks (accounting, marketing, etc.) required to make the company into a viable system.

Note that, by virtue of Ashby's Law, the total variety on the vertical plane (including System Two) must equal the sum of the variety on the horizontal plane. The set of Systems One, after all, constitute the operating unit for a System One at the next level of recursion. All the variety in this operational unit has to be accommodated (albeit with attenuators and amplifiers) in its communication with the manager.

Viewing the viable system in its present configuration, this presents a problem for our earlier assertion that there should be minimal intervention by System Three into the activities of Systems One. Rather, the variety equation requires considerable intervention, almost to the extent that meta-management is taking on the responsibility of the System One managers. Figure 5.12 shows this with its emphasis on the vertical communications.

Figure 5.12. Horizontal and Vertical Variety.

There is an alternative, however, which is shown in Figure 5.13. The idea here is that the meta-management can directly interrogate the operations. Some authors and managers call this “management by walking around”. It provides important information to meta-management, which allows it to convey much less information down the management channel. For example, drivers are given a set of rules to follow in driving, so that it is not necessary for the state to control each trip. This direct communication allows monitoring of the operation to ensure that it is running well. If it is, communication with Systems One is limited. This in turn allows the variety equation to be satisfied, even as it reduces the amount of direct intervention by System Three into the System One Managements.

Figure 5.13. Direct Information.

Note that care must be taken to design these direct communications between System Three and operations, so that they do not interfere with the smooth operations of System One.

How do I manage a multibillion-dollar company in a world where the business environment changes almost daily? And the problem isn't even that there is insufficient information about what is going on. It is that there is too much information. Not only is there too much information about what is going on within our organizations, there is way, way too much information about what is going on in the environment.

Variety has always been what management has been about, and it has grown much worse in modern times. To say that we are suffering from “future shock” is to say that the variety of our environment has exploded.

What this means for systems designers is that our number-one job is to create the amplifiers and filters necessary to enable a manager to survive. For filters, we must design reports that provide enough information, richly enough, to be useful in managing the enterprise, without providing so much that the important items cannot easily be identified.

Vice presidents, for example, are not going to go through a two-inch-thick report to find the four pieces of information they need. On the other hand, in 2001, a two-page summary of Enron's financial status clearly wasn't adequate to understand what was wrong at that company.

For amplifiers, we must set up structures such that a single command from above can easily be translated into a complex set of instructions to all involved.

Moreover, as we design these systems, we must understand which system we are serving: Systems One, System Two, System Three, System Four, or System Five.

And what does this model tell us about the movement toward knowledge management?

Knowledge is the application of information to useful ends. That is, knowledge is the variety that has been filtered to equal the variety of the manager perceiving it.

The meaning of knowledge management, then, is twofold:

First, the ability to absorb variety is now the most important trait anyone brings to the job of being manager. This capacity for dealing with variety is a direct function of intelligence and education. To deal with all that is going on around the enterprise, managers simply have to work smarter than they ever did in the past.

Second, to the extent that the world still will always have more variety than the manager can absorb, information systems are required that digest the information and produce the specific sets of information that allow for ready interpretation. To the extent that our information systems simply add to the noise, they provide no value to the enterprise.

To the extent that they provide exactly what is needed, they are indispensable.

Traditionally, enterprise organization has emphasized the relationships between Systems Three and One. Systems Two and Four, if they existed at all, were diffused throughout the enterprise.

The cybernetic systems tended to look like this.

Now, however, technology permits much more interaction among all the players. The management role at System Three doesn't disappear, but many more people can participate in it, as well as Systems Two and Four.

Now you get something like the following configuration, in terms of the four twenty-first century jobs described previously (see page 209):

In addition to the issues of variety that are present in managing organizations, similar issues are present as we manage our personal lives. Each one of us (forget about the enterprise for a moment) must suddenly deal with innovation, changes in technology, and a vast amount of new things that must be learned—this week.

But we all have a limited “channel capacity”, and we each deal with this fact in our own way. We select which magazines we don't read and, of those we do, which articles (or paragraphs) in them we don't pay attention to. We carefully don't hear discussions about subjects that are troubling or that might challenge our preconceptions. Similarly, we amplify our knowledge in the way we use those tidbits in conversation to give the impression that we know more than we do.^[12]

As individuals, we destroy the variety we experience as well, and we amplify the variety at our disposal.

The assignment is to somehow organize our lives so that the gleaning of tidbits is systematic and at least leaves us aware of the things we are supposed to know, even if we don't quite know them yet. We also try to broaden our education as a way to increase channel capacity.

Of course we go to conferences and seminars and the occasional class to try to improve at least our intellectual tools for dealing with variety, if not our knowledge directly.

And we don't always succeed. There are religious wars about the best data modeling techniques to use, because none of us command enough variety to comprehend the full array of techniques that are out there and the implications of each.

Perhaps understanding this model will allow us all to accept our limitations and get on with our lives.

A very important part of the documentation for a data model is identification of the roles (if not the individual people) who will be responsible for each entity and attribute. Certain people will “own” an entity, meaning that those people are ultimately responsible for its data quality. The same or other people are responsible for updating it. Yet others have permission to see it. This matrix is required if data quality is to be established.

Documentation of each entity should at least include identification of the role responsible for the quality of its information, if not the individual person so responsible.

All of the activity modeling techniques (except for the function hierarchy) described in Chapter 4 include references to the people or organizations performing the functions. Data-flow diagraming and process modeling were described in Chapter 4, because these techniques emphasize what is being done; use cases were described in this chapter because they are more concerned with the nature of the interaction. Indeed, all three of these techniques belong firmly in both columns.

In addition to their basic definitions in the models, all activities should also be classified as System One, Two, Three, Four, or Five activities.

Where people are significantly influences the way Systems One through Five described here work. Each person is in a particular location, and the roles each plays may be for different locations. It is incumbent on management of an organization to place the people who must communicate most extensively in the same location. The extent to which they are in disparate places will affect the quality and nature of the communications channels that can be built among them.

In the cybernetic model of the organization presented here, the main topic is communications, with no distinction being made between simple data and events. One event that was significant, however, was an operation's going outside its operating values (set points). Ultimately, this is the event that requires the management entity to act. With all the events that management must respond to from the outside world, this internal event is also critical.

The analysis project must clarify which events affect which parts of the organization.

As stated previously, Chapter 8 contains a model of an organization's ends. It specifically deals with what it calls “elements of guidance” (business policies and business rules) that amplify management decisions, and “assessments” that provide a filtered view of the effect of the environment (there described as “influences”) and operations on the organization's ends and means.

An element of guidance represents a message sent by management to its operating divisions. Business policies and business rules must be carefully designed to account for the issues described here.

Similarly, an assessment is the retrieval of information from the operating organizations, interpreting what happened, which influences caused it to happen, and the effects of these events on the enterprise's ends and means.

Business rules may be specific to the organization involved. There may be good and valid reasons for a rule to be applied to one division and not another. This must be documented.