Questions about edge work

by Philip Boxer

The following questions were asked by Larry Hirschhorn about the blog on empowering the edge role, following which are my responses:

Larry: You say that edge work is breaking out almost everywhere. Can you give me an example?

Philip: We worked recently with the members of a clinic providing through-life care to patients with chronic conditions. The hospital managed the clinic’s activities in terms of treatment episodes, and budgeted the clinic according to their annual episode throughput. Because patients stayed with them long term, their budget tended to keep increasing because more patients were referred to them than died or left the area. But their budget couldn’t keep increasing. So instead patient care was rationed, the result being systematic under-use of the treatments available, to the long term detriment of the patients.

Larry: Where does ‘edge’ come into that?

Philip: As part of an acute care system, it made sense to treat the clinic as part of the centralised drive to provide cheap treatments. In these terms, the patients had to fit in with the hospital’s programming to provide the most efficient use of its resources. But the conditions in the clinic needed to be treated on a through-life basis, each course of treatment being designed to fit the particular patient’s condition. This meant the clinic’s (and behind that the hospital’s) resources fitting in with the programming of the patient’s treatment.

Larry: What made the project challenging?

Philip: First of all, although the clinicians had high situational awareness of their patients’ conditions, their ability to synchronise the behaviour of their own and others’ clinics around the needs of the patient were very constrained by the hospital’s internal processes of referral. Secondly, there was no way of holding the clinicians accountable for the way they aligned treatments (and costs) to patients’ conditions. Thirdly, the senior management used fixed budgets as a way of controlling the aggregate performance of clinics, there being no way of flexing them according to the mix of conditions different clinics were being asked to treat.

Larry: And how do you relate these challenges to the failings of bureaucracy?

Philip: Bureaucracy works by taking a big problem and breaking it down into little problems. The assumption is that if you solve all the little problems, the big problem will be solved too. This is called ‘deconfliction’ – each little problem is defined in such a way that it can be solved independently of the other problems. So the clinics with their associated budgets were defined on this basis, and as long as the levels of demand were stable and there had been enough time for patterns of treatment to settle down, things could work okay. This might have been true when both treatments and conditions were simpler, and the medicine less advanced. But it is not true now. The increasingly chronic nature of conditions, combined with the need for treatments aligned to the particular patient, means that treatments have to be adapted to patients and not vice versa. Bureaucracy can’t cope with this.

Larry: So what did you have to do differently?

Philip: There were a number of things. Firstly, we had to alter the way clinicians made referrals to each other so that authority to determine treatments was delegated as close as possible to the edge. Then we had to support the way the clinicians needed to manage treatments in a way that could also hold them accountable for the performance of those treatments. Thirdly, we had to change the way budgets were allocated so that they were based on the mix of patient conditions being treated. And finally, we had to secure funding on the basis of the cumulative cost of securing long term patient health outcomes instead of on the basis of the short term incidence of acute conditions.

Larry: Were you successful?

Philip: We proved that it was much cheaper in the long run to organise the clinic in this way in 6 different hospitals. But to sustain the changes it meant reallocating budgets on a different ‘edge’ logic across clinics. This the larger systems weren’t ready to do.

Larry: Would you say that it created more effective government due to edge work?

Philip: More effective government requires three things: the political will to do things differently, the ability to create value at the point of contact with the citizen/patient, and infrastructures sufficiently agile to make change practicable. We failed on the third count. Our brief was to make the edge work better, not to change the agility of the infrastructure. And I think that is the particular difficulty. To make the edge work better, you have to change the agility of the infrastructure, and the vested interests of the former are very different to those of the latter.

Taking power to the edge by empowering the edge role

by Philip Boxer

An edge role is on a task boundary in which the systems of meaning on either side of the boundary are different: some form of translation is required. The task facing a multi-disciplinary team is often such that its members are in edge roles with respect to the areas of specialism they represent. In such a team what is being but into question across any given edge is the identity of the specialism. The challenge for the individual in this situation can be expressed in terms of taking up such a role within the life of the enterprise. The normal assumption is that, if only the nature of the task itself was sufficiently deeply understood, then ways could be found to reach an accommodation between the differing systems of meaning in the interests of the task as a whole. This is, for me, the implication of identifying the problematics of the ‘mid-game’ with those facing the edge role.

But when the team is the enterprise itself, even if such an accommodation can be found between its members, there is a further challenge: the challenge of the particular nature of the demand being presented by the client. Now the team as a whole is on an edge between the system of meaning implicit in the way they work together, and the system of meaning required to satisfy the demand. So the experts may agree how to treat the patient, but the patient may not survive the long term effects of the treatment – the patient needs to play a different game to the one the experts want to play.

The relationship to demand
Two dimensions are proposed:

  • the first describing the form of the synchronisation of task activities needed to deliver the effect required by the demand. Here the distinction is between the ‘logic’ in terms of which the form of this synchronisation is defined. High situational awareness is needed if the form of synchronisation depends on the nature of the demand.
  • The second dimension is the dynamic relationship required between the timing and logistics of what is done and the demand situation itself. Here the distinction is between whether or not the two are coupled. With coupling comes the need to manage the synchronisation with the situation itself.

So we get the four characterizations of bureaucrat, intelligence analyst, SWAT team and edge worker. But what happens when we look at the way the enterprise is able to sustain such different forms of relationship to demand?

The double diamond
The diamond above, describing the relationship to demand, is the right hand diamond in this diagram. The left-hand diamond describes the way the enterprise sustains relationships to demand. Thus the two diamonds are symmetrical around the central vertical axis, and if some intervention creates a lack of symmetry between the positions on the left and right, then there will need to be some form of process of adjustment to bring them back into alignment. What is on the left side?

The alignment of the infrastructures

  • The first corresponding dimension is the form of responsibility given to the individual for how task activities can be synchronised. The distinction is between doing this through a span-of-control – essentially vertical relationships; or through a span-of-complexity – horizontal relationships. The span-of-control has built into it the means that imply the end, while span-of-complexity has to build means from assumptions about the end.
  • The second dimension is the form of accountability through which those responsible are held to account for the performance of the enterprise in relation to the demand. Here the distinction is between accountability in relation to hierarchy or to situation. From the point of view of hierarchy, means-ends relationships created under its aegis will by definition be transparent (regardless of what others might say were their appropriateness), while those created in response to situation will not be (because they cut across the boundaries defined by the hierarchy, and require horizontal forms of transparency).

So we get the same placing of bureaucrat, intelligence analyst, SWAT team and edge worker. And we can read ‘football player’ for SWAT team. But what about the positioning of the counter-insurgency player and of the chess player?

Who are we trying to talk to?
The chess player I have argued really belongs in the intelligence analyst position because of the underlying bounded nature of the problem space. But need the counter-insurgency position be in the intelligence analyst position?

If we contrast a ‘hearts-and-minds’ (HM) approach to counter-insurgency with a ‘helicopter gunship’ (HG) approach, then the HM counter-insurgency player will be in the edge role position, but the HG approach uses intelligence to find out where the baddies are, and then sends a SWAT force out to zap them… so what would bring someone employing the HG approach to take on the HM problematics facing the edge worker? The answer has to be based on its effectiveness in response to asymmetric forms of threat, although behind that answer there is clearly an issue of how power is used – an ethical issue associated with the long-term effects of using overwhelming force.

Ironically, the architect might be a better metaphor for the edge role position, because the form of the structure being asked for is much more arbitrary, and the challenges of aligning the available infrastructures in creating a satisfying response to the demand very evident. The architect faces the challenge of enabling the enterprise to take up a role in the life of the individual demand.

The elevator conversation
There was an article in the FT today about a study by Gartner surveying 1,400 CIOs:

CIOs had noted a shift from back-office functions towards front-office projects such as improving customer data management, looking to help the business stand out with strategic and innovative use of information, business processes, and intelligence in products and services.

To enable the business to stand out in this way is to enable it to take up edge roles… every business has to work out how to face this challenge in their own particular way.

DoJ search requests

by Richard Veryard
cross posted from SOAPbox blog

Various people talking about the dispute between Google and the US Department of Justice (DoJ).

Boing Boing, Search Engine Watch, Emergent Chaos, Graham Shevlin.

One line of discussion relates to privacy. Some commentators are praising Google for resisting the US Government’s demands for data, when its competitors have apparently complied.

But there is another important consideration, which appears relevant to Google’s position. There is an huge gap (asymmetry) between the information requirement (as stated by the DoJ) and the data on Google’s database. (See especially Danny Sullivan’s post at SearchEngineWatch.)

So the DoJ is hoping to solve a complex problem with very large amounts of data? Does it really make sense for the DoJ to copy the raw data from Google onto its own processors? In a service-oriented grid-enabled world, it would seem to make more sense (and raise fewer privacy concerns as well) for the DoJ to collaborate with Google (and its competitors) – to compose intelligent and relevant analytical enquiries that can be run by Google (as a service, albeit commandeered by the Government) to help solve the DoJ’s problem.

Of course Google is an interested party in the outcome of the DoJ’s deliberations, but does engaging it as a trusted partner in the analysis really increase its ability to bias the outcome in a self-interested way? And if it provides knowledge and metadata services rather than raw data, this might mitigate the threat to its position as a trusted custodian of personal search records?

Banking Services and User-Defined Policy 2

by Richard Veryard
Who is going to want the kind of user-defined policies I talked about in the podcast (link to soundfile, transcript extract)? Is it just the higher-end type of customer, as Ron suggests?

Hypothesis One: The better-off customers have the more complex requirements. Their financial arrangements are more subtle, their consumer electronics are more sophisticated, and there is much greater scope for interoperability.

Hypothesis Two: The technically literate consumers are most able to articulate the more complex requirements. They are more willing to experiment with the available options, and to learn to express their requirements in an appropriate policy language. In consumer electronics, they are the ones who know the difference between an Ethernet cable and a USB port, and how to tweak the firewall.

Hypothesis Three: The lead times are getting shorter. Even if it is currently the better off and technically literate consumers who are the early adopters of this complexity, service providers should anticipate the possible mass adoption of some aspects of this complexity within a fairly short timescale.

Hypothesis Four: Although the better-off and technically literate customers may be the ones who currently understand and express these complex requirements, that doesn’t mean that the rest of the customers don’t have these requirements. Everyone needs security; and the less money you have to start with, the more you suffer if someone steals a hundred dollars from your account.

Hypothesis Five. Where end-users are not able or willing to engage with the technical necessities (such as writing their own policy statements in some technical mark-up language), there will be intermediate services that will do this. For example, financial advisers may start to see their role as helping the client to orchestrate and manage a complex set of financial services from a range of service providers, instead of simply helping to select financial products. There are also opportunities for self-help groups and communities to emerge, where the complexity is managed collectively at group level, rather than at the individual level.

Hypothesis Six. Ultimately, the complexity is supported by a platform of composable services providing the right balance of flexibility and efficiency. The strategic question now becomes one of platform dominance. (The banks may be privately thinking about this question, but I haven’t seen much evidence of it yet.)

Hypothesis Seven: In the short-term, banks might be tempted to focus on the higher-end type of customer if they really were the most profitable. But there may not be enough of them to cover the costs of supporting them effectively. A more strategic reason for focusing on the higher-end type of customer might be because of innovation. But there may be just as much innovation (and greater social benefit, as well as reasonable long-term profitability) from supporting a “long tail” of lower-end customers, either directly or through appropriate communities and intermediaries.

Banking Services and User-Defined Policy

by Richard Veryard
Transcript from Podcast [34:30-38:38]

[Richard] Let me talk about the relationship I have with my bank. As a typical banking customer, I get a very simple set of services from my bank, and quite frankly it’s not really what I want, but all the other banks offer me pretty much the same services, so I don’t really have much choice. What I would like from my bank would be for me to define my own data model, which would be more complex and more fitted to what I need than the data model the bank gives me. What I would like is to be able to define my own policies on the bank account – give me a policy language, I’ll code my own policies, I’m happy to do that – and let the bank execute them. But no, the bank’s not interested in doing that, the bank can make money just giving me a standard one-size-fits-all bank account, and so they’re not going to do that, and they’re probably not going to do that for some years to come. But other industries are starting to respond to that kind of asymmetric demand.

[Ron] That sounds really interesting about the bank, but I can’t even imagine or conceive of how a bank could possibly allow me to define my own policies, or the kind of data that’s going to be related to those policies. Can you give me an example of something that might work that way?

[Richard] Yes, let me talk about security policies. At the moment, my bank gives me a simple choice: either everybody with my password can access my internet bank account, or my bank account simply isn’t available over the internet. And so it’s a very crude binary: either it’s open or it’s closed. Now what I would like is to define much more precise security policies on my account, that says for example I can take money out of my account to these specific destinations up to these amounts of money, but if I want to pay vast amounts of money to an overseas company that I’ve never dealt with before, I do not want that to happen over the internet, I’m quite happy to go into the branch and do that over the counter, and sign all the paper that I need to give myself the extra security. I could write those policies over and above the policies the bank itself has, and the bank would be able to execute my security policies in composition with its own security policies, and that would give me greater security and me greater control over my account, without taking anything away from the bank. And if everybody had their own security policies it would make it a lot harder to have mass attacks on bank accounts, which would make everybody safer. And so there seems to be a win-win-win all round if banks were able to provide just simple kinds of user-defined policies of that kind.

[Ron] Okay, I can imagine how if that kind of thing were possible, that a bank who was willing to offer that might attract the kind of customers that would result in more profitability. I would imagine that might appeal to a higher-end type of customer than a typical customer that maybe doesn’t care that much about the bank’s security.

[Richard] But you see with SOA, the technology is all there to do that. Technologically, that’s very easy to do now. It’s purely a question of whether the bank is willing to manage that additional complexity.

For further discussion, see Banking Services and User-Defined Policies 2, in which I discuss who is going to want these kind of user-defined policies, and what are the strategic implications for banks and other service providers?

Architecture Podcast

by Richard Veryard
A 40-minute podcast Philip and I recorded with Ron Jacobs in December has been released on Microsoft Channel Nine.

0:00 Introduction
1:58 Governance Ron
2:32 Asymmetry Philip
3:39 Value of SOA to business – the adaptive enterprise Richard
5:25 Asymmetry – Medical Example Philip
6:44 Silence
7:07 Impetus for SOA – take costs and risks out, put value in Philip
7:47 Fluctuations in demand volume Ron
8:06 Implications for the relationship between service provider and consumer – does the service provider know what the fluctuation means for the consumer – efficiency versus flexibility Richard
9:44 Implications of interoperability in the real-time enterprise – supply chain example Philip
10:49 Interoperability from demand-side perspective – joined-up-government example Richard
13:18 Governance at the edge of the organization – asymmetric governance Philip
15:32 Metropolis and city planning as metaphor for SOA governance – Christopher Alexander and the Nature of Order Richard
17:48 4-colour workshop – Blue, Red and White – the role of Enterprise Architecture Richard
19:02 City planning example – city planners as enterprise architects Ron
20:14 4-colour workshop – the role of the Black team – anticipating the challenges of the response to demand Philip
22:44 Enterprise architecture that fails to deliver adaptability. Problems with universal models. Richard
24:22 So who does the black team thinking? Examples from public sector, military, telecoms industry, voluntary sector. Philip
26:41 Anticipation of context in which demand emerges Philip
27:30 Internet business models – Pareto thinking versus “long tail” thinking Richard
29:26 Paying attention to customers one-by-one – customer context – school example Philip, Ron
30:25 Whose priorities – provider’s interest versus consumer’s interest – school example Richard
31:33 Diversity of interest – patient records example. Philip
32:12 User defined meaning – and technorati Richard
32:42 Consumer-side flexibility – agile systems Philip
34:30 Banking services and user-defined policy Richard, Ron
38:38 Close Ron

Type III Agility in Organisations

by Philip Boxer

The best metaphor for type III agility comes from understanding the way air superiority fighters work. Early aircraft were designed to be stable in flight: the aircraft designer knew how they would fly before they took off. But air superiority fighters are designed to be unstable.

To understand how this works, we need to introduce the idea of variable geometry, for example by having a variable-sweep wing. By putting a computer between the pilot and all of the control surfaces available to the aircraft, the computer can work out what is the best way of keeping the aircraft in the air and responding to the pilot’s requests. The computer has to be able to do this within the physics of how the variable geometry makes it possible for the aircraft to fly. When it can control thrust as well, the computer gives the aircraft supermaneuverability. Thus we can say that what makes these aircraft different is the fact that they have to work out how they will fly as they go.


So if we now return to the concept of agile organisation, we must first describe organisation in terms of how it imposes particular geometries on the relationships between the individual capabilities of the organisation – geometries-of-use. Thus for example, the business process re-engineering of value chains is used to create competitively optimal geometries in the way an enterprise creates value.

If we look at military organisations, we see them developing such concepts as network enabled capability (NEC), and network-centric operations (NCO). We see these innovation happening elsewhere too – for example in telecommunications in response to the needs of corporate clients, or in healthcare where treatments have to adapt continuously to a patient’s emerging needs through the life of their condition. The aim of these innovations is to make the geometries of their organisations variable – variable geometry organisation. This is what creates Type III Agility, needed to meet the challenge of the 21st Century.

What is at stake is the ability of organisations to organise themselves around the needs of their customers, instead of requiring their customers to organise themselves around what the organisation is able to provide. Of course this raises the question of who is the ‘pilot’, which for agile organisations is the question of how to take power to the edge.

The 21st Century Challenge to leadership

by Philip Boxer
The Tavistock Institute of Human Relations recently published this paper on the challenge faced by leadership in the 21st Century in their pamphlet on learning (page 2).

The paper argues that this challenge is a double challenge: organisations are being presented increasingly with forms of demand that are asymmetric to their assumptions about demand. Even if customers don’t want more, they at least want what they are being offered in a way that relates particularly to them. This means that for leaders it is not sufficient to find new ways of addressing demand – it is also necessary to question the existing terms within which they themselves understand demand. Asymmetric demands require you to pay attention to what you don’t know.

The lesson to be learnt from the military about threats is that you can die if you ignore them. In creating a double challenge for leadership, the threat presented by asymmetric forms of demand is that they question the current basis of power and authority of the leadership itself. This is the 21st Century challenge to leadership – to find the basis of their authority in the value they create within their customers’ contexts-of-use.

The Tavistock Institute is an example of an organisation committed to taking up this double challenge in its work with its clients.

BlockBusters and SlowBurners

by Richard Veryard
On the Longtail blog, Chris Anderson produces some data that he interprets as evidence for The Decade the Blockbuster Died. Many comments (including mine) suggest alternative interpretations of this evidence.

An album (or film or other entertainment product) may achieve massive volumes when it is first released, or it may become a perennial best-seller. In theory, there are two routes to long-term success, so that a slowburning quality product might achieve parity with a flash-in-the-pan popular hit. But in practice, products are deleted from the catalogue or become effectively unavailable – either because the sales volumes falls below some threshold defined by economies of scale, or because of some arbitrary decision by the record company (such as a dispute with the artist). In the former case, such thresholds assume an economics based on there being a symmetry between the supply model and the levels of demand directly addressable by it.

So the statistics cited on the Longtail blog don’t reflect the “pure” behaviour of the market but reveal the distortions caused by the behaviour of the record companies. This is an example of large companies imposing arbitrary restrictions to suppress the longtail.

In recent years, the internet has apparently provided opportunities for record companies to exploit the back catalogue more effectively, but there are still many classic records and films that are not available in the latest formats. Pundits or rigged elections produce lists of the 100 greatest products in various categories (special thanks to Channel Four) and this promotes sales of these older products. But the market remains distorted, and the evidence for the longtail remains patchy.

Similar considerations apply in the pharmaceutical industry. But in the same way, a business model reliant on blockbuster drugs is looking increasingly problematic, while the industry has not yet made a satisfactory transition to the longtail.

The strategic challenges in these (and other) industries can be understood in terms of meeting the challenge of asymmetries of demand, and the need, therefore, for a business model that can use a micro-segmentation of demand to address the needs of customers one-by-one.

Type III Agility

by Philip Boxer

In the previous blog on the three agilities, a list of benefits identified by a Gartner report on agility (“The Age of Agility”, 2002) was organised into three groups corresponding to three types of agility. Why? And what does this tell us about what is required to deliver type III agility?

Two underlying issues are being addressed here: firstly, whether or not there is more than one autonomous supplier involved with the user; and secondly whether or not the demand from the user has been anticipated by the supplier. This gives us the following 4 quadrants:


Thus the two approaches to satisfying anticipated user demands are ‘directed’, either by the single supplier, or by several suppliers who choose to come together under a single contractual arrangement. The ‘Type I+’ position is anomalous here, since if there is only one supplying organisation and it encounters an unanticipated demand that it can satisfy from its existing capabilities in time, then it will do so on a contingency/ad hoc basis using directed composition. If it cannot, then if it is to respond to the demand, it will have to collaborate with others in order to bring in the missing capabilities. This will face it with the challenges of the Type III quadrant, where a collaboration of suppliers who do have the requisite capabilities must agree how to organise themselves to meet an unanticipated demand in a way that is specific to the demand.
So what is required to deliver Type III agility? Mark Maier, in his paper on Architecting Principles for Systems of Systems, distinguishes between directed and collaborative systems of systems (SoS). In the diagram above, his directed SoS correspond to Types I and II, while his collaborative SoS correspond to Type III, virtual SoS being large-scale versions of collaborative SoS. (Maier’s examples are the world-wide web for collaborative SoS and national economies for virtual SoS, but see separating the supply-side from the demand-side).

Thus collaborative SoS involve separating a supply-side infrastructure (Maier’s examples of these are the internet itself or industry structures) from demand-side organisations that are free to organise supply-side services in whatever way they choose, hence distributed collaborations.  This is the separation that presents the supply-side infrastructure with asymmetric demand. Type III agility is therefore characterised by the need for distributed collaborative composition described by Richard in asymmetric design – collaborative relationships between suppliers, a joint appreciation of what is driving demand, and a capability for collaborative composition. What distinguishes distributed collaboration is therefore that the identity of the collaboration is emergent in response to the demand itself. This involves taking power to the edge.