Why is it so hard to prove the value of generalist disciplines such as enterprise-architecture? And what can we do in practice to resolve this challenge?
As I see it at present, there are three distinct yet interrelated parts to this, which I’ll split over three posts:
- part 1 (this post): about the underlying assumptions that give us so much trouble
- part 2: about structures and interdependencies between services
- part 3: about where and why Taylorism is a poor fit for many present-day business-needs – and some key implications that arise from that fact
First, though, a quick round-up of where we’ve gotten to so far:
- ‘On learning enterprise-architecture‘ – where all of this discussion started, exploring the skills needed for generalist disciplines such as enterprise-architecture
- ‘There are no jobs for generalists‘ – about why it’s so hard to get work as an enterprise-architect, and why most advertised ‘jobs’ for enterprise-architecture don’t actually have much to do with enterprise-architecture itself
- ‘More on ‘No jobs for generalists’‘ – exploring the implications arising from various responses to the ‘No jobs for generalists’ post
You’ll probably get the best value from this series of posts if you’re already familiar with some of my other work on whole-enterprise architectures, such as Enterprise Canvas service-oriented modelling, the Market-Model / Market-Cycle model of enterprise dynamics, the SCAN sensemaking/decision-making framework, and the notion of ‘variety-weather‘.
The ultimate roots for all the problems we’ll see here lie somewhere in Taylorism – so-called ‘scientific management’ – and, behind that, linear concepts of causality and ‘control’. You’ll probably need to understand some of those ideas, too.
It’s not essential to know all of this beforehand, though, and anyway I’ll explain the key points from those frameworks as we go along.
Two paradigms – or three, perhaps
There are two approaches or paradigms to business – or to anything, really – that we could summarise respectively as linear or flow. As we’ll see later, a systemic paradigm is a kind of superset of both of those two paradigms that aims to create context-appropriate syntheses between them.
The linear paradigm asserts that everything connected by strict patterns of cause-and-effect. We could also describe this as a digital paradigm – a world of explicit boundaries and sharp-edged cut-offs between ‘true or false’. Everything can be calculated exactly, and to exact precision. Sometimes there may be delays and other factors that may affect the overall calculation somewhat, but in essence, once we know what the parameters in each context are, and how to change those parameters, we can control that context.
[In essence, Taylorism is the direct outcome of the linear-paradigm applied to business at large scale.]
Everything is distinct, separate and, perhaps most all, certain. And because of this, we can make control easier by breaking things apart into smaller and smaller chunks – in other words, ‘divide and rule’ – and then working on each of the problems within each of those chunks or ‘boxes’ as a separate, distinct entity.
Each box is a discrete, defined unit, a component. These components have discrete interfaces: to create a business-process, we chain components together in a predefined sequence. To change the functionality of the whole, we pull the components apart, and reassemble them in some different configuration. Again, divide-and-rule, but also linked to recombine-and-rule.
A specialist is someone who works on one specific type of box, with specific skills that relate to the functions and parameters that apply to that box. In a business, work is described in terms of distinct and explicitly-defined specialisms, tasks and responsibilities – in other words, a ‘job-description’ – that would apply to a specific box.
Boxes can be described at different levels of granularity: in essence, boxes nested with boxes within other larger boxes. A silo in business-terms is, in effect, a large box or cluster of boxes that all relate to a specific area of expertise.
In this view of business, seniority is related directly to the size and scope of the box: the more boxes contained within that box, and the number of layers of other boxes within that box: the more contained boxes, and/or layers of boxes, the more senior you are. All decision-making flows ‘downward’, or ‘inward’: outer boxes determine the activities of inner boxes – never the other way round. A C-level executive is ‘chief’ or ‘controller’ of a top-level box, and all other boxes within that box; the CEO or Chief Executive Officer is the ‘controller’ of the metaphoric box that contains all of the other boxes.
Thinking of the (business) world in this way – a world of strict causality, of strict certainty – does make things easy to understand, and especially to describe to others. Anywhere in the mid-range of physics and the like will also appear to conform to this paradigm of ‘scientific law’ – which is one reason why most machines, and almost all computer-based ICT-systems, are designed at present around these types of principles.
It’s a type of thinking that, if social-mapping with MBTI (Myers-Briggs Type-Indicator) is correct, would come naturally to perhaps 70-80% of people – in other words, very much the majority. However, that also suggests that thinking in terms of any other paradigm may well be a lot harder for the majority of people – which is a point that many of us may perhaps forget…
The flow paradigm takes the opposite view: everything is in flow, everything is in flux, always undergoing change. We could also describe it as an analogue view, where no sharp edges can exist, no black-or-white distinctions but everything in shades of grey – or all the colours of the rainbow. Properties emerge within a context that exist only between things, not in things – hence trying to describe a context only in terms of ‘things’ makes no sense. At its most extreme, this view asserts that there is no causal connection between anything and anything-else, in fact no ‘things’ to be causal about: instead, everything just is. To use the ‘boxes and lines’ metaphor, there are no boxes, there are only lines – or even no lines either.
There is no certainty; there are no states, no components; there are no rules, in fact the whole notion of ‘divide-and-rule’ makes no sense, because there is nothing to divide – or if we did, it would cease to be what it is.
It’s literally anarchic, Without some kind of focus, though, it does have a nasty tendency to drop back to what we might call ‘kiddies’-anarchy’, a random ‘me-first’ or even ‘me-only’ mess.
It might at first seem that there’s no place anywhere in business for this view, and that nothing could be achieved with it. But in fact it’s at the core of ideas such as Agile, or almost all forms of innovation – anything that deals with change as change. The challenge though, is that it can often be very difficult to make to ‘make sense’ of anything from this view, simply because everything is changing; and likewise even harder to describe to others, other than through phrases such as ‘the Tao that can be told is not the eternal Tao’, which may technically be correct but are not exactly helpful…
[There’s probably no equivalent of Taylorism here, in part because the flow-paradigm is so hard to describe and explain to others.]
It’s a type of thinking that MBTI social-mapping suggests would come naturally to perhaps 10-20% of people – very much a minority, yet still a significant presence. Again, however, thinking in terms of any other paradigm may well hard for such people – hence the ‘culture-clash’ described in CP Snow’s classic The Two Cultures, for example.
A systemic paradigm tries to find a balance between those two extremes: it asserts that since both are true, in effect neither can be true on its own, and instead there needs to be some kind of synthesis – both boxes and lines – within a broader systemic scope.
It’s a type of thinking that MBTI suggests would come naturally to only about 5-10% of people. In other words, an even smaller minority again – but a minority within which we’d find most full-scope generalists, such as futurists, enterprise-architects, and designers of almost all kinds. If we do fit within this minority-group, one of the hardest things for us to remember is that most other people do not think this way – and if we want them to understand what we’re on about, it’s up to us to do the appropriate translation into their modes of thinking.
The catch, of course, is that the synthesis is neither truly linear or ‘digital’ – which leaves the linear-paradigm folks unhappy – nor truly free-flowing ‘analogue’ – which leaves the flow-paradigm folks unhappy. To make things worse, the balance is context-dependent: it sort-of stays still for a while, and then changes again, in line with what we might call the ‘variety-weather’.
In short, just about everyone can be unhappy with it, if they want. But unfortunately it seems likely that it’s the only way that architectures and the like can actually work…
This applies especially once we start to take real-world dynamics into account. In essence, the Taylorist-style notion of boxes-within-boxes only makes sense if the world is entirely static – which it isn’t. Things change – which means the boxes get distorted, and no longer fit, or else straddle uncomfortably across two or more silos, or whatever. The more we try to control the hierarchical structure of boxes, the more brittle the whole thing gets – and if there’s nothing to take the strain, the only feasible end-result is a complete collapse into a puddle of broken boxes. Not exactly the outcome we would want for a viable business… So the boxes alone are not enough: we need something that will take the strain of change.
A real software example of what’s needed here is Donald Knuth’s typesetting system TEX. There, the core metaphor is ‘boxes and glue’: each letter, word, paragraph, text-block and so on is a ‘box’, linked to other boxes by invisible blobs of ‘glue’ of varying degrees of ‘stickiness’ – which means that we can stretch the spacing between the boxes, and shift the relations between the boxes, in any way that we need, constrained somewhat by the stickiness of the glue. The same is true in the self-reconfiguring ‘responsive’-type layout of this weblog, as you’ll see if you resize the browser-window: columns change width automatically, and the widgets on the right-hand side will relocate themselves to the bottom if there isn’t enough room for them otherwise on the page. Boxes-and-lines, or boxes-and-glue: they’re very similar metaphors, about how things hold together and relate with each other in dynamic ways.
Implications for recruitment
In the Taylorist model of business, every activity can be broken into predefinable tasks, and the capabilities required to implement each task into a predefined box called a ‘job-description’. Everything is defined in terms of objects, as ‘resources’: hence people as ‘human resources‘, interchangeable units of capability than can be plugged into tasks as required – and discarded, like any other resource, when worn out or no longer needed.
[Many of us might think of this as inhuman, or worse: but note that it does make sense within its own terms, if we think of the world solely in terms of externalised machine-like objects. Remember that a logic cannot assess the validity of its own logic: the same is true of a paradigm, but on a much larger scale. To assess whether a paradigm is valid within a context, we have to be able to move outside of the paradigm: but the whole point of a paradigm is that it purports to be ‘the answer to everything’ – that there is nothing ‘outside of’ that paradigm. Failure to grasp the implications of this is a huge source of problems in enterprise-architecture and the like…]
In the Taylorist view, there are only boxes: the task of the (external) analyst and/or manager is to shuffle the boxes into whatever configuration might be required for the current task. There is no glue between the boxes: the analyst/consultant is in some ways the nearest equivalent to glue, but all of that not-really-glue is outside of the overall containing-box.
Every box has its ‘job-description’; and every box represents some form of specialism. There are no generalists in Taylorism: every box is complete in itself, and all the boxes fit together with ‘scientific’ precision. The general notion of a ‘job’, as we know it today, is almost entirely an artefact of Taylorism. Of which a direct corollary is that since Taylorism has almost no concept of a generalist, and certainly insistent that it should have no need for one, there are no ‘jobs’ for generalists.
There’s one further twist we need to note: that at root, Taylorism pursues an almost-mythic notion of ‘scientific’ perfection. It asserts that, in theory, everything in the world should fit into a specific configuration of boxes: our task, as practitioners of ‘scientific management’, is to identify that perfect configuration. If we find some task that seems to straddle box-boundaries, all we need do is re-identify those boundaries. If we cannot do so, we have failed Taylorism – not the other way round. In other words, like many (perhaps all?) paradigms, Taylorism here contains within itself a classic circular-proof – of which we definitely need to be aware…
In principle, Taylorism should work: it’s certain that perfection can be achieved. As The Mikado puts it, “My object all sublime / I will achieve in time / to make the punishment fit the crime / the punishment fit the crime”: the aim of Taylorism is much the same in business. In reality, though, such ‘perfection’ could only be achieved in practice if the world stayed perfectly still and static – which it doesn’t. The very act of changing something changes the system itself: hence the very real need for some kind of glue to take up the load. Otherwise known, in business, as the generalists.
So at first glance, a flow model of business might seem a better fit for a dynamic business-world, because it asserts that everyone is a generalist. In practice, though, every ‘job-description’ would look like this:
- work as directed, if you feel like it
- make it up as you go along
…which might appeal to many, but it doesn’t fit well with any kind of collaborative work. In short, a flow-model notion of a ‘job’ does not scale, because, almost by definition, everyone would be wandering off in their own direction, following the flow. And pure generalists are often so intent on the following the flow that, to be blunt, they just don’t get things done – or, perhaps more to the point, completed to a form that’s usable by anyone else. Which again isn’t much use in business.
Which brings us to the obvious need for some kind of notion of ‘a job’ – some systemic synthesis – that’s somewhere between those two extremes. In a systemic model of business, we still have a somewhat-Taylorist notion of ‘jobs’ in boxes – but the boundaries of the boxes are somewhat ‘squishy’. We need that ‘squishiness’ in order to allow for movement, for flex, for limited change; but we also need definite in-the-box skills and capabilities in order to get things done. That’s the whole point about specialism: it’s not so much the specialism itself that counts, but the ability to deal with the fine-detail that applies at the actual moment of action.
And in addition to the boxes, we need to pay explicit attention to the ‘glue’: we need explicit generalists. Yet, unlike the unconstrained flow-model generalist, the way we’d typically describe those generalists is in terms of the types of boxes which they link together. In a sense, it’s somewhat like the Taylorist notion of the ‘containing-box’, except that it doesn’t assume an exact fit between all of the contained boxes. Instead, it’s more like the skull, the brain-box, or containing-musculature for inner organs in a boundary: it either changes shape itself or, if it must be mostly-rigid – like the skull – it’ll contain some kind of fluid to dampen out any shocks or strains. Hence, for example, a domain-architect: specialist in the sense of focussing on a single domain, yet also connecting across all of the different views of the true specialists within that domain.
Yet the systemic view also has to take into account one other key factor: the need for skill. In the terms of the SCAN framework, skill sits on the far side of the Inverse-Einstein Test: it’s how we deal with ambiguity, with inherent-uncertainty. In the classic Taylorist view, there should never be any need for skill: everything should be automatable, so skill is something we should always aim to eliminate, just as we would aim to eliminate all forms of complexity, and replace it with perhaps-complicated but always-controllable algorithms. In Taylorism, as in the Newtonian science that underpins it, there is no such thing as inherent-uncertainty: it doesn’t and cannot exist – because if it does exist, we couldn’t control it. (Which as we’ll see in a later post in this series, is perhaps the real issue at stake here…)
But since Reality Department – and more modern science, for matter – does demonstrate that uncertainty is inescapable, so too will we always need human skill to work with it. Which leads to a very important point:
- the real-world always includes some degree of inherent-uncertainty
- automation depends on rules, and on certainty
- machines and (most) IT-systems operate only on rules – they cannot cope with ambiguity
- some IT-systems can simulate some types of types of skill via sheer speed, but internally still operate only on rules
- skill is the human mechanism via which we adapt our ‘rules’ to cope with complexity and inherent-uncertainty
- despite the deep desires of Taylorists, no system of automation can ever be complete within itself – it will always require the support of human skill to adapt its rules to changing needs and changing contexts
Which leads to a very dangerous point indeed: the more we automate, the more we put at risk the means via we which we cope with the inherent limitations of that automation. (For more on this point, see the Sidewise post ‘Where have all the good skills gone?‘) The more that Taylorism ‘succeeds’, in that sense, the more it sows the seeds for its own inescapable destruction. And the pace of change we face in business now requires ever-increasing adaptability – which is a skill in itself.
Which tells us, in turn, that recruiting by ‘job-description’ alone is never going to be enough: in the terms we’ve used above, recruitment must describe both the box and the glue. Every ‘job’ will include some distinct mix of specialist and generalist – never just specialism alone. And, in effect, every ‘job’ must describe how it will support the learning that will make that ‘job’ viable in real-world practice – otherwise it will all but guarantee business-failure, especially in the longer term.
But if that’s what’s required in recruiting, where do people acquire the respective skills? Which is where we hit up against yet another Taylorist-inspired problem: the all-too-common inability to understand the crucial distinction between training and education.
Implications for business-education
Taylorism likes training. Training is all about creating a mechanical process that can deliver the same response to the same input, time after time, in a predictable, controllable, certain manner. Most of our culture’s so-called ‘education’-systems are actually training-systems in this sense: it’s about rote-learning, repeating exactly what the teacher taught, and nothing more. That’s how you get it right. (Or, perhaps more importantly for many, about how you don’t get it ‘wrong’.)
In short, it’s how we teach people to be machines. Except machines are better at being machines than people are. And once we’ve taught a machine what to do, it’ll keep on doing the same thing over and over until someone tells it to stop. Hence the endless desire for automation: replace expensive, unreliable, unpredictable people with cheap, reliable, predictable machines – and at last gain complete control over every aspect of the business.
Except it doesn’t work like that: no system of automation can ever be complete within itself. Which is where skills come into the picture.
And skills cannot be taught: they can only be learnt. Whilst training is about pushing rules in, education is about bringing skills out.
The next catch is that skills are personal. They’re not exchangeable objects, or computer-programs that can be swapped in and out at will: they arise only from synaptic-connections and much else beside that occur only within that person. If we want to create some kind of ‘skills-transfer’, the best we can do is provide under which others can learn. In effect, the task of education is to provide conditions under which skills can be learnt.
Next is that it takes time to learn new skills. (See the Sidewise post ‘10, 100, 1000, 10000‘ for more detail on this.) And the skills-learning process is never straightforward, always confusing, always error-prone – especially in any kind of social context. (See the Sidewise post ‘Surviving the skills-learning labyrinth‘ for more detail on that.) Again, skills arepersonal: each person takes their own time, and learns in their own way. Trying to force them to do it the same way as others, or at others’ pace, will only slow things down. It isn’t training: it’s a fundamentally different process.
There are important meta-skills, too: for example, learning how computer-languages work in general, rather than solely learning a single computer-language that could be out of date in a couple of months. (When – as all too often – a recruiter sends out a job-description demanding ’10 years experience’ in some computer-package that has only been available for a year or two, what they really mean is ‘some experience in that package, as part of perhaps ten years experience in some vaguely-related field where the meta-skills would match’: and it would help everyone a lot if they actually said that in the first place…)
And since every employer will make use of these hard-earned skills, we should regard continuous-education as an explicit employer-responsibility – not something that they can somehow handball everyone else. (Anything less is actually a form of theft – either from the community or other employers, or literally life-theft from the employee.) Which means that continuous skills-education needs to be part of every job-description – and every employment-contract, too.
There’s one other crucial point we’ll need to note here. As I said in an earlier post, the fundamental present-day concept of a ‘job’ is the outcome of a Taylorist fantasy: people as interchangeable ‘assets’, as ‘units of productive capability’, components in the organisation-as-machine. In large part it stems from the Marxist notion of ‘excess-value’, the value that could be extracted from people’s labour in the context of capital’s control of ‘the means of production’. At the time when Taylorism really took root, around the start of the 20th century, the primary ‘means of production’ was in factories – places such as Ford’s assembly-lines, or, before them, William Blake’s ‘dark satanic mills – and also still in agricultural land or down the mines. In a knowledge-economy, though – as is much more the case in the present-day – the ‘means of production’ reside primarily in individuals’ skills and knowledge, much more so than in machines themselves. In that sense, the Marxist model is inverted: which means, again, that our recruitment-models are ludicrously-outdated for present-day needs.
The core of a knowledge-economy is people’s individual skills and knowledge: and in practice, by the very nature of skill and knowledge, those can only be possessed by that individual – not by some external ‘owner’. It’s still common in business to talk about as people as assets – such as in the well-meant phrase “Our people are our greatest asset!” and suchlike – but in legal fact the only the time that people themselves literally are ‘assets’ is when they’re slaves. This is not a good way to describe people if you want them to commit themselves to your enterprise… Instead, the respective asset is the organisation’s relationship with that person – not the person as-asset. And maintaining such relationship-assets – without which the organisation has no access to that person’s skill and knowledge – requires a fundamentally-different approach to recruitment, employment and engagement than in the ‘person-as-component’ concept used within a Taylorist-style concept of the architecture of the enterprise. For ‘jobs’ that involve any form of skilled-work, the relationship is the ‘asset’ – not the person: we forget this fact at our peril.
Enough for now, anyway. In the next part of this series I’ll describe how a service-oriented view of enterprise-architectures can help to clarify how the ‘boxes-and-glue’ concept actually works in real-world practice. And in the final part I’ll explore how Taylorist-thinking still pervades most current business-school ‘education’ – and how what it teaches is, for the most part, exactly how not to run a modern corporation.
In the meantime, though, any comments on this so far? Over to you, if you would.