 All right, it's a real pleasure to have Dave Snowden over here. It's a dream come true for me I've been trying to get Dave Snowden maybe for six years now and finally we have him here I first saw his keynote in 2007 in Limerick, Ireland and Probably till date I think that was one of the best keynotes I've ever seen I was literally at the edge of my seat throughout the keynote I'm sure you're gonna get the same experience today. He speaks really fast So you have to be attentive don't do multitasking while he's speaking you will get his jokes ten minutes later other ways All right with that. I'm gonna call Dave Snowden. Thanks a lot for coming Dave Limerick was interesting. It was the only the second agile conference I'm done in fact it was the first because I did an xp event in London and Having spent most of my early life in it. I was one of the founders of DSDM Etc and then moved into general management it was rather odd to come back in and Discover that the Canadian frame which you designed for leadership development and corporate strategy had been picked up in my old Area of development IIT So for the last four or five years I've been having a thoroughly enjoyable time of plunging back into project management in large IT projects and doing things I thought I'd abandoned early on in my career and coming back So you should hopefully get a bit more reflective wisdom now than you did in Limerick The general title for what the stuff I do and other people do is called sense-making There are various definitions of that. I define it as how do we make sense of the world so that we can act in it Now with that comes the concept of sufficiency yet. You can never know everything you need to know So how do you know when you know enough and what sort of actions can you do based on what you actually know? And really everything falls back to that as we move through and one of the key lessons of this And this is really important for those of you who are focusing on simplistic recipes And this is from gaping void. If you don't know gaping void brilliant cartoonist You can go into his website and one of these cartoons comes down into your email every day If it wasn't for that and Dilbert, I think I'd go quietly insane. All right, they keep me going But this makes a really important point all paths up are different all paths down to the same So the idea that somebody can give you a recipe which will mean that you will succeed is just fundamentally flawed All the people I hear speaking at keynotes whether they're companies talking about changing markets or people talking about Radical new methods for agile have actually not copied what somebody else has done They've done something radically different and they've changed the space Yeah, and that's really important to understand. Yeah, I can give you a recipe for failure And the recipe for pay for failure is repeating past mistakes Come back to something. I said briefly on day one Yeah, people are arguing that some of the scalable frameworks and necessary because it's the only way to get adoption of agile within the company Well, if you get people to adopt agile without understanding agile, you're just dooming yourself to another IT failure Yeah, it may give you a temporary satisfaction of suddenly getting budget But downstream the expectations won't match the delivery and you're back into a more traditional mold of IT So there are times when you have to get it right up front and maturity and wisdom is about knowing when that's the case I'm not taking a simplistic route Am I gonna make a big difference between being simplistic and being simple and that that's an important distinction So what I want to do is to introduce some of the basic concepts around complexity I'll do round that around the Kinevin framework. Can I just do a quick check? How many people have heard the children's party story? Oh Good a naive audience, right? Okay, you'll get those who have you'll get a live performance, all right So now I know I can use that one After that I'm gonna go through two areas of development where we're working now actively within cognitive edge one is Looking at the whole area of how we change the area of requirements capture and how we move into a continuous feedback loop between requirements capture software testing and Really moving to a much earlier stage of understanding users If we go back to some of you know, just great stuff yesterday about the number of failures that you see I want to be slightly different on that. I think okay, so 80% or 90% of what we do fails I've heard that from people who've got lots of experience. I don't want to just accept that I actually think we need to move learning much earlier into the cycle I think the reason we have that amount of failure is we wait too long before we start actually interacting with users and development So I don't want to take those failure figures. I want to talk about learning early to fail less Yeah, and I'm going to talk about different methods for engagement for that Now I'm going to give you a very brief insight but not much detail into some of the new complexity approaches We're now adopting to large-scale IT project management Which actually recognized the value of waterfall? I famously once said that safe is Prince 2 in agile clothing With a deliberate reference to wolves and sheeps. All right Fundamentally, I'm now working with the Prince 2 community because we actually think we need to start to resurrect some of that material But actually understands its interaction with agile because there's a large amount of IT project management We're somewhere with the discipline of Prince actually has application Yeah, it's just we need to know its boundaries and that concept of boundaries is going to be key to what I talk about If people understand that you understand boundaries You're using the right methods in the right places Then you're going to scale to all sorts of levels without a problem If you're going with missionary statements about if we just adopt this and have everybody certified in the next six months Life the universe and everything will be wonderful. Then sorry you're going to fail again Yeah differentiated approaches are key So let's move on to Kenevin Kenevin is a Welsh word It's pronounced Kenevin It means in Welsh. It doesn't mean habitat. That's an English phrase The English don't have an equivalent Many of you will know I'm Welsh so defining ourselves as not being English is key. All right That's part of our national psyche In Welsh it means the place of your multiple belongings. It's a sense of being rooted in many different pasts Which profound the influence what you are, but of which you can only ever be partially aware So it's a very good name for a complexity model because that's the reality of life. You're in a flow of meaning over time Yeah, at any one point in time you can act in that time But you do it hugely influenced by factors from the past of which you can never be aware That's true of your users. It's true of your C-level executives. It's true of everything. Everything is about flow It's not about static assessment And there's going to be a lot of lessons about moving to flow dynamics and away from static linear approaches So Kenevin works off some basic science What myself and my own group in IBM decided to do almost 20 years ago now is that social science was in a very bad place Now I knew that already my first degree is a joint honours between physics and chemistry Sorry, between physics and philosophy So I was taught to despise social science from two completely different disciplines and that really hasn't gone away since So we decided to go back to complexity science to cognitive neuroscience to the biological end of anthropology And say what do we now know about systems and people in decision-making and let's actually Reapply that into social systems Rather than actually try and build on what is actually a fairly perverse discipline which makes wrong assumptions So that's where it's comes from So within within that basic science of systems we can identify three different types of system Effectively ordered systems complex systems and chaotic systems Now I should be clear at this point that there is no full conformity of use of language in the field Yeah, physicists use complexity and chaos differently from social scientists different from chemists different from biologists So if anybody wants to say that's not chaos that's randomness. Yeah, okay. I agree. I'm not gonna get fussed about that Yeah, in the disciplines. I come from that's what we call chaos other people call it randomness I'm gonna define my terms, right? It will take another 20 years before the field settles down. Yeah, that's the nature of new things So I use the constraint-based definition which comes from Gerard and others and as further developed by myself And basically what I'm arguing is that a highly constrained system is what we call ordered The level of the constraints are such that there are no freedom of movement. You can't actually escape the constraints They contain all behavior Now that type of system is unique to humans. We're really good at it and it has huge value These days we count the number of surgical instruments left at the end of an operation and do a checklist procedure to check It's the same as we're there at the start Yeah, as I get older, I think this sort of thing is of increasing importance I know the figures for the percentage of surgical instruments left in people's bodies before that was instituted and it's a really scary Percentage So we're good at that sort of stuff, but we don't understand that it's deeply contextual Highly disciplined control checklist procedures only seem to work for human beings in highly ritualized environments So in an operating theater the process of scrubbing up changes the way you think Now I've been into an operating system theater as an observer twice Yeah, both my children were to quote Shakespeare from their mother's tomb and timely wet ripped And I've now decided that Caesareans are wonderful from a father's perspective You get to see a fascinating operation. You get to see the baby first and you get to hold it first So I'm all in favor of them, right? But that process of going through washing your hands cleaning yourself putting on the white boots and everything You actually think differently at the end of that process and that's actually what ritual does in human systems A ritual entry into a highly highly role-based process Means that you will follow rules that you would never follow as an individual person And you can see that massive compliance with hygiene in operating theaters Low compliance with hygiene in wards where people aren't ritualized into the role And there's a body work I'm not going to talk about today that we're doing for example to ritualize entry from development into software testing Literally if you get people to change their clothes, you will improve their behavior as a tester Because actually that ritual entry by changing a costume changes the cognitive activation patterns of the brain Now I say there's a lot of depth in this I'm giving you some hints of things But order has value but realize the context The danger is that we apply it where it shouldn't apply and one of the many advantages of working for IBM for seven years I hastened to add I didn't volunteer I was conscripted and I was the only person in the history of IBM to last seven years without filling out a timesheet I'm still proud of that achievement. All right was an active continuous petty defiance. All right, but it mattered Is that you learn everything there is to know about excessively constrained system? I mean there's another line heuristic in IBM don't buck the process There's no point in arguing the process is wrong from a customer or a business perspective because you'll be killed The only way to actually achieve customer satisfaction is to work around the rules In fact the whole secret of survival in IBM and I doubt if it's changed since I left is the ability to manage within a rule constraint system, which in many ways is perverse So for example when they banned us buying food for staff late at night Yeah, which they said was in you know, we didn't need to do it And if we did need to do it, why didn't we ask in advance? Well, if you could ask in advance You wouldn't have to be buying food at four o'clock in the morning because the system's got to go live at 9 a.m Yeah, they eventually made it impossible. So the practice emerged of overtipping London taxi drivers So you've got a blank receipt Filling out the blank receipt for the amount of food. Yeah submitting that receipt and getting your money back Now I made this reference at scrum the Scrum Alliance conference in billion last year three people from IBM ran up and said We're still doing that. Did you invent it? And Somebody from HR came up to me and I'm new from HR and IBM And she said we knew you were doing that, but it was easier to let you carry on with it now It won't take you long in any large organization to see similar patterns The obsession with order the obsession with making things look superficially structured Means that actually the informal networks have to work beneath the surface to make apparently Efficient apparently efficient systems effective So the over focus on efficiency the over focus on process Produces perverse results and you can see that the same sort of thing in a lot of agile methods We start off with an inspirational set of ideas yet from snowbird and it ends up getting to a structured logical ordered linear process that people aren't allowed to escape from and Then those structured ordered linear processes a link with other structured low loaded processes the point where we get diagrams with So many bloody boxes and arrows Yeah, the intention is to create dependency on the model creator rather than to create something people can actually use Yeah, so order has value, but within limits and the limits are actually much narrower than we think Chaotic systems on the other hand I'm defining the systems without constraints, which isn't the same thing as randomness All right, I'm still trying to find the right word for this an unconstrained system is one where everything is operating Independently of everything else in the context in which you're working Now that can be a catastrophic failure Yeah, all of a sudden all the old uncertainties all the old certainties disappear What will actually that's only ever a transitionary state if you ever have a crisis Within minutes people start to impose constraints And a big mistake people make when they use Kinevin is to assume chaos is a permanent state actually it isn't it's a transitionary state Yeah, that's very important to realize it's a transitionary domain Used deliberately it has huge power. So for example, if you can remove all constraints, you can get huge novel innovations But it takes a lot of energy to create a system without constraints Yeah constraints are natural to human systems. They'll come in place immediately So if you genuinely want to make them thing truly innovative You have to put a lot of management time and energy into keeping things apart and keeping them separate Otherwise they'll get killed The other big use and this is something we're working on heavily at the moment is a variation of what's traditionally called wisdom of crowds Which is the ability to do whole of workforce engagement in real-time decision support because if everybody assesses something independently of everybody else and you look at the results graphically on the screen You may recognize some lean concepts here. Yeah is is mass engagement But visual representation to allow human beings to make decisions Fundamentally that allows us to actually do real-time decision support under conditions of uncertainty We're now starting experiments within Wales and potentially New Zealand and Nova Scotia What's called a small countries project to actually create citizen networks by which we can ask questions of entire populations in real-time to inform government policy So once you start to understand the context of and lack of constraints There's lots of fascinating things you can do with it, but it takes a lot of energy and a lot of structure And then we move to complex adaptive systems. This is the day-to-day reality of most life These are systems where you have what are called enabling constraints Constraints which actually to a part extent modify behavior, but behavior themselves modifies the constraints So what actually happens is the constraints and behavior co-evolve key phrase from biology Yeah, as people interact with constraints and act interact with each other interact with tools interact with systems interact with their environment The constraints are modified the people are modified the systems are modified things are in constant flux or constant change And one of the key things about co-evolution is it's associated with a concept of irreversibility You can never go backwards and Anybody with teenage children knows about co-evolution and irreversibility You can't actually say let's actually do a level set. All right. Yeah, we got a real problem here, right? They're 14 or 15 anybody know the problem when they hit 14 or 15 Yeah, and there's things you really wish you hadn't done about 15 years ago. It was kind of like that was a mistake I now realize it You can't say well, let's get in some management consultants and let's do a strategic reappraisal of our child rearing strategy Let's set key performance indicators milestone targets You know that let's run a series of linear iterations to see if we can get this behavior, right? You don't do that. Do you? You got to be you're gonna need to be scared about this by the way I was in America recently and this when I satirized this and discovered there are now consultants who will actually come into your family and Create mission statements value statements KPI's and learning objectives. So This actually gets quite depressing. All right when you think about it Now the reality is complexity is about inherent and continuous Uncertainty, but it can be managed as a flow. It can't be managed as a static system Yeah, we can all manage flow Yeah, if you can ride a bicycle, you know if you try and slow down too much It's a problem if you speed up too much. It's a problem, but provided you keep the right speed. It's easier to maintain balance Yeah, so think about ways in which you manage dynamic flow and that's what complexity is about Now the key lesson on complexity And this is something which is difficult for people to get is A complex system is not causal if you get that I don't need to do anything else But that's a really difficult for people trained in a Western tradition to understand It's actually much easier from a Catholic or a Hindu or a Taoist tradition Because those religions have kept the concept of some things just are Yeah, there's no reason for them to be they just are Yeah, whereas a Protestant tradition has got inherently causality built into it The basic fact is a complex adaptive system has dispositions At any one stage in time it may it's more likely to move in one direction than another direction But it might perversely move in a completely unexpected direction. So it has dispositions, but not causality Yeah, the same thing will only happen again the same way twice by accident not by the inherent nature of the system Technically for the philosophies amongst you it doesn't have linear material causality Yeah, but that's kind of like the most common understanding of causality. Yeah in most most management So you can't say if I do this I will get that result Complexity theory effectively a priori invalidates any method which says it can produce a defined result Yeah, and that actually changes the game If you have to manage with systems, which is dispositional you have to manage in a radically different way You have to manage the evolutionary potential of the present moment in time and adjust as you go Something hinted at by quite a few of the speakers yesterday I now put in the theory on what people with practice already understand. This is actually a theory to back up common sense Now the best way I've ever learned to understand the difference is is to think about organizing a party for a bunch of Nine or ten year old children if you can all imagine that imagine this and you're making the mistake of holding it in your own house This is an error right you learn this pretty fast All right the reason of church halls and community centers Yeah, a better venues for parties is they have fire hoses And fire hose is a very useful for cleanup after the party and they're occasionally needed for crowd control during the event itself All right, so yeah, I recommend the church hall So let's look at how we'd manage the children's party dependent on what type of system it is So if we assume the system is chaotic that means the children are unconstrained their behavior is random Which means they're probably discover drugs and alcohol and go on a personal experience of self-discovery Your house may burn down in the process, but all property is theft and it was socially constructed in the first place So why are you worried about it? And there's a stato in that for some academic colleagues of mine. I don't recommend this method I've got friends in California. You've tried it, but never more than once. Yeah, the recovery cost is very high The order systems approach on the other hand You'll all understand if you've been on any of the popular certification courses Under this you agree learning objectives to the party in advance of the party itself You make sure that learning objectives are aligned with the mission statement for education in the society to which you belong and It clearly articulated and printed off on motivational posters with pictures of eagle soaring over valleys and water dropping into ponds And you put the posters around the room where you're going to hold the party You then produce a project plan for the party The project plan should have clear milestones throughout the party against which you can measure progress against ideal party outcome The senior art our part senior adult should start the party with a motivational videotape Yeah, you don't want the children wasting time in play which isn't aligned with the learning objectives of the party itself And then the said senior adult should use PowerPoint to demonstrate their personal commitment to the party objectives And to demonstrate how the children's allowances are linked to the achievement of the milestone targets Following the highly successful completion of the party you conduct the after-action review Yeah You then identify improvements to best practice regimes you mandate process improvements and everything is wonderful If any for wrote for it for any remote reason the children aren't happy Then you impire an appreciative inquiry practitioner will get until happy-clappy stories So they have happy mental models and suitably indoctrinated. They're like whatever you put in front of them next time Everybody reasonably familiar with this approach to yeah party management Yeah, it's a scalable approach to party management. I strongly recommend it On the other hand the complex systems approach is much simpler We start off by drawing a line in the sand known as a boundary in complexity theory And we look the children squarely in the eye and say cross that you little bastards and you die One of the things you learn for pretty fast as an adult is the value of flexible negotiable boundaries Because rigid boundaries have a habit of becoming brittle and breaking catastrophically We then introduce catalytic probes a football the video a barbecue I won't say cricket ball because I understand national grief at the moment It's any comfort we lost those bastards English it wrote me this year as well So I'm in a bad mood on that one. It's the same sort of phenomenon Fundamentally what we do. I shouldn't go these down the versions I we introduce things that kids may play with and sometimes they play with them and sometimes they don't if they do And it's a good thing. It's called an attractor and we amplify it if they do and it's bad thing It's still called an attractor, but we do our best to disrupt it very quickly That's where you use the fire hoses. All right, I remember the fire hoses. They're really important What we're actually doing it effectively is managing the emergence. This is a key phrase beneficial coherence within attractors within boundaries and Actually that phrase in the sequence is very important. You can't actually create an attractor. You can only catalyze it But if you have catalyzed one, it's working you amplify it if it's not you disrupt it and Boundaries you can manage in a more structured way Now I make no apology for using specialized language Yeah, Heidi go famously said man thinks he's a master of language, but language is the master of man We know from a cognitive neuroscience point of view if you don't change language, you don't change the way that people think So people who refuse to learn new language have actually refused to change or to think and that's very basic Now too far it's jargon, but some change of language is key So I'll try and keep it to a minimum, but the concept of catalysts attractors and boundaries is vital Now you should start to see why complexity is a huge importance to government We're now as I say starting experiments with a new center for applied complexity at the University of Bangor in Wales To actually do whole of country experiments This complexity offers us an alternative between the extremes of free market capitalism and state planning It actually says if we put state resources into managing boundaries and creating attractors locally contextual solutions can emerge at Significantly lower cost than either of the current mechanisms Yeah, the cost of free market capitalism is massive in terms of social deprivation and inequality The cost of state planning is massive in terms of bureaucracy There has to be a different method and complexity gives us a science-based approach But what I say about countries applies to companies as well Yeah, complexity is a new approach to governance with multiple ramifications It's a genuine paradigm shift in the way we think about the world and it requires people to think differently But therefore to act differently in terms of the way it works that lead us on to Kinevin And in Kinevin What we actually do is drawn like this. All right, so please do not draw it as a two by two matrix and call it quadrants All right, it's a five domain model. You can't have five quadrants. All right, this is basic language All right Or don't draw it as a two by two matrix and put a diamond in the middle because you know I'll pick it up on Twitter and get irritated. All right, sorry I'm going about existing practice. It's drawn with it with five strokes of the pen And what it does is devise ordering to two obvious and complicated Now this is an important distinction Because in obvious order I'm everybody can see the relationship between cause and effect as a result to which everybody will go along with best practice This is the only valid demand. It's not just that there's a linear relationship between cause and effect Because the high level of constraint produces predictability, but I can apply rigid constraints Because everybody buys into the solution And that's really important to understand because the personal buying is key to this working My favorite example of this is you all know this is civilized countries drive on the left-hand side of the road Sweden decided in the 60s to stop being civilized and become uncivilized and they all Moved over from the left to the right. It was quite hysterical All right, I remember seeing it when I was a kid on the television But it was the right thing to do because all the countries around them Yeah drove on the right now. I realize in Italy and India. This doesn't apply. All right There's a whole different driver behavior in in Italy and India is actually mathematically predictable based on flocking flocking algorithms So, you know that there are different models that you can work from all right But the point is you know in the UK we decided to drive on the left now once you've decided to drive on the left That's it All right, everybody can see it. It's reasonable. You should do it. Of course. There's some deviation Yeah insisting you drive on the left when actually if you don't go on to the right-hand side of the road You're going to run over a kid is actually ridiculous So even in rigidly constrained systems, you should always allow for exceptions Something people forget in process improvement exercises But in the vast majority of cases you can apply best practice We then move on to complicated where again, there's a linear relationship between cause and effect, but fundamentally here It's not self-evident to everybody. So we have to do analysis or we have to bring in expertise Which means we have to trust the experts and part of the problem with trust in experts is we asked them to make judgments in Complex domains where they're always going to get it wrong or they're only going to get it right by accident So we lose confidence So one of the other reasons to make the complex complicated distinction is to know when you can trust experts And when you can't in terms of their overall judgment of outcome Yeah, it's actually about building better trust in many ways Yeah, so here I do analysis I bring in experts I sense analyze respond and now I have governing constraints not rigid constraints Governing my constraints means there are degrees of freedom if you rough the right expertise and experience Yeah, even in this space over constraint can produce perverse outcomes And one of the big problems with consultancy led initiatives is consultants like to create neat tidy diagrams Yeah, and they actually like to force people to select between options The reality is when you work with experts and I did a huge amount of work in knowledge management back in the 90s Basically experts don't know what they know until they need to know it So interviewing experts and then mandating processes based on what they remember at the time Means you get things radically wrong in the future and again you're building over constraint So allowing a degree of freedom is necessary in the complicated space, but again It's then based on qualification experience peer review. There's all sorts of checks The basic point about both ordered domains is you've got effectively an external skeleton Yeah, within that skeleton you've got freedom with rigid order. It becomes very tight with governing constraints It's looser, but it's an external constraint With complexity, it's very different It's now more like the internal skeleton of a mammal than the external skeleton of an insect And remember internal skeletons allow huge amounts of variety Yeah, and actually far more adaptable Now in terms of individual species not in terms of overall evolution, but in individual species So it's kind of like they're called enabling constraints because they allow a leave evolution to happen And one of the key applications of this in agile by the way is the concept of self-organizing teams If you just allow people to assemble with whoever they want, it's actually chaos It's actually doesn't really produce good results if you introduce constraints about team organization and team membership Those are called enabling constraints They provide enough structure that the system will evolve in a favorable way And again, this is theory-informed practice once I understand the basis is a complexity I know without constraints the thing is random. Therefore, I don't get any reliability So the issue is what type of constraints do I introduce and when do I do them? Don't underestimate the power of constraints Well again anybody with children knows about constraints Yeah, yeah, the issue is you have to know when to use the constraints and to be honest The ones that they don't realize are visible are the most effective Interesting human beings habit is a more effective constraint than rules Yeah, and actually building habitual behavior is a more effective intervention than designing rules, but that's for another day So basically we then move on from that to a chaotic system in a chaotic system We have an absence of constraints which again we can do novel practice. That's the basic Kinevin framework We also have the central domain of disorder which everybody forgets Which is a state of not knowing which of the systems you're in It's not the same thing as chaos because actually in disorder you could be ordered. You just don't know it Right some of the effects may be the same because disorder and chaos are always transitionary in Kinevin The only stable states are obvious complicated and complex So using it as a categorization model makes a mistake, and I've seen some major errors on this people say my method is complex Yours is complicated Some scrum people are very prone to this. You know scrum is complex. Therefore, it's good Well, actually there's nothing good or bad that they've been complex or complicated And actually the great power of scrum is it's a complex to complicated transition device It actually makes complex things complicated. That's its power Yet moves things across the boundary And the danger is people use the domains to privilege things. You know, you're ordered. Therefore, I can dismiss you We've now got a new group. All right one systems architects amongst them We're actually saying that chaos is the really valuable space and they can manage it because they're wizards I love that phrase Everybody in complexes complexity is now passe and there's desire to privilege different domains is actually very dangerous All the domains have value And it's the ability to move between them is key so there are some basic dynamics on Kinevin and People really need to sort of present these as well This is the basic cadence the basic dynamic as things are complex. We see patterns patterns emerge We stabilize the patterns as we stabilize them yet. We can actually shift them into the complicated domain So the basic principle of complexity based intervention is you start off with multiple parallel safe to fail experiments Which is why crumb scrum is not a true complexity technique Because it does one thing in a linear way This is my force things earlier on in the cycle and we actually call this a pre scrum technique within agile You do smaller experiments faster Yeah in parallel I'll talk later about our new work on narrative if you see a cluster of user narratives Which actually achieve statistical significance and let's take a technique from XP You put a pair programming team on to those stories and produce a small prototype Now that's a pre add scrum technique and you can do multiple ones of those in parallel The ones which gain traction then move into a scrum So you're moving from the center of the complex domain into the boundary Once you're in the boundary then you use scrum to move it across the boundary Start to get the principle here That dynamic is key because actually what you do in a scrum is you iterate and see if it works as the user accept it You're testing if the constraint is replicable So once the constraint starts to produce repeated behavior, you know, it's complicated now You can scale it in the traditional sense of the word If actually it doesn't continue to produce repetition then you need to move it back into the complex Yeah, so that dynamic between complex and complicated is easy to manage It's a natural cadence and you need to work it out. What it is for your industry sector. It might be days It might be weeks. It might be months. It might be years, but it's going to be different If you don't get that right, you might have to do what's called a shallow dive into chaos Yeah, that's a radical disruption because you've got negative pattern entrainment Only a small amount of things go down there This is actually again the big danger with over codifier methods It takes 10 or 15 years for any method to consolidate You don't over codify until you've done enough experiments Yeah, and that's why I actually quite like Prince 2 because its basis has 10 of about 30 40 years behind it So modifying that is actually more effective than trying to start again I never thought I'd be advocating Prince 2, but life has moved on So basically you only put a very small amount of stuff down into the obvious space because anything which goes down there Isn't going to need to change again. That's the key principle right over over a sustainable period There's actually a new dynamic, which is becoming more important in software We call this the grazing dynamic. You won't see this in any of the papers yet Which actually says some things stay in the complex permanently yet they graze order and they graze chaos and And we're now starting to see you can see this in social computing But we're starting to see it in object-based architectures is the ability for a system never to stabilize but constantly to deliver value And I haven't got time to go into this I've given a few lectures on on complexity approaches to software architecture But that creates architects architectures which allow things to remain permanently in the complex domain Yeah, and actually small code constantly are assembling with defined input output gives you huge value there Now if anybody wants to know more about that contact me because to be honest, that's the future Yeah, the level and the speed of change in society means we're going to be spending more time in the complex with less Opportunities to move into the complicated than traditionally we've had but our architecture and development methods don't currently permit it But we're fairly close The basic metaphor behind all of this and this is from a Gerardo is that actually with organizations We're dealing with bramble bushes in a thicket or a mango mango swamp if you want something which is more local Yeah, you know, there are separate plans, but you can't see how they're distinct Everything is entangled everything is connected. You pull one thing unexpected things happen. Does that sound familiar? Yeah, well, that's actually where we are with software code at the moment Yeah, that's why neat tidy structures don't work. You're dealing with bramble bushes in a thicket You've got legacy systems unstated requirements constant change Everything you do affects everything else. All right, you're in bramble bushes in the thicket. That's the metaphor. It's an ecological metaphor It's not an engineering metaphor Which actually is key because if you switch software to an ecological concept It's very different from an engineering diagram and that that's kind of like the big change So some things to give you a hint up. This is a basic diagram Diane will recognize this it's moved on a bit since this first version This is starting to think about project management from a complexity perspective So on the bottom line we have waterfall development Yeah, there's a huge bound to the large software project where everybody knows what they've got to do And they should just go and do it Right, so that's water fall at the bottom. We then move into complicated That tends to be scrums or sprints You've got lots of these working in parallel because you kind of like know what the user wants and you're pretty close to it Yeah, therefore you're moving across Then we move into the complex where you're doing multiple small groups of parallel Experiments much earlier on in the cycle Yeah, and then in the chaos that's an intervention mechanism So you can see here we're starting to show a time-based pattern in which the multi experiments feed into scrums But then if the scrum starts to go wrong, let's suppose you go through four or five iterations Yeah, and the user still isn't happy you trigger it back into the complex and Trying to create a method by which people will do what scrum was originally designed to do which is sometimes not go through to completion Yeah, and actually it's very rare for people to do that We need that that greater dynamics if you get an alert in the obvious space it moves immediately into crisis management in chaos So something goes wrong in complicated you shift it back to complex something goes wrong in in Rigid-ordered systems in the obvious system It's a crisis because you thought you had complete predictability again We're working on this at the moment, but this is a new dynamic visual representation I'm working also with David Anderson at the moment because we're working on a complexity approach to Kanban Which will be a non-linear representation of progress First course on that is in London shortly because we're gonna to be honest It's easier to work out a new approach by teaching it. Yeah, so that that's what's actually going on We're combining Kanban with complexity and creating new methods of representation complexity I'll show you one of those in a minute to give you an idea Now part of all of this and I'm gonna flip a bit from complexity now and talk about cognitive science This is one of the most scary things Because if you read this properly the systems analyst profession needs to collapse and close down There's huge amounts of experiments on this 24 radiologists asked to look for a effectively a cancer nodule Yeah, in one of the x-rays they were shown a gorilla picture, which was 48 times larger than the cancer nodule was present Yeah, and as it says 83% didn't see the gorilla even though the rise scandal Now this is scary The basic facts are the most you will scan of what in front of you is three or four percent And that's on a good day five percent at most You then match it against previous patterns, which are half remembered patterns of past failure Here directly or here through narrative and you do a first-fit pattern match The way we make decisions is we scan stuff through our recent memories and based on a partial data scan We associate with those recent memories and we make decisions Anybody who conducts any more than three interviews has actually formed a subconscious Hypothesis and they literally only pay attention to things that match that hypothesis Now that means you can't trust an analyst to go and interview users and have any chance of understanding what they need Not only that the users don't know what they need because they've been interviewed in a specific context start to see why we get massive failure Because at the best we're doing partial scans Now in evolutionary terms this makes a lot of sense. You can't afford to scan everything and make rational decisions Interestingly the only people who do that are autistic and Just to scare you the two university departments with the highest degree of partial autism a computer science and finance Right, so that has some implications as well if you think about it because it's a positive advantage We we actually do scan more from an IT perspective than users And then we assume that users will see things the same way we do. We're actually they don't Know very different perspective in evolutionary terms You can imagine it think about the early hominoids on the savannahs of Africa Something large and yellow with very sharp teeth runs towards you at high speed Yeah, do you want to artistically scan all available data? Then look up a catalogue of the flora and fauna of the African belt I haven't identified the lion Look up best practice case studies on how to deal with lions Yeah, by that time the only document of any use to you will be the escape manual from the digestive tract of a large carnivore and The only example I've ever found of that is the book of Jonah in the Old Testament and I don't recommend building management practice based on that Yeah, we evolved to make decisions very quickly based on recognition of past patterns So what do you think happens when you launch a new initiative to the IT community? We're gonna put you all through training courses We're gonna certify you how many times you've done that to them before So what are they filtering it through? You go to users and say I know we balls up the last five systems, but now it's right because we've adopted a new technique Don't worry What are they scanning it through? You're always dealing with people's perception of the past in the present They will not actually listen to you if you talk about a future So actually doing small things in the present is actually more effective That means we got a look at the world in a very different way Again, one of the things we focused on is what's called scalable ethnography Now this doesn't mean the sort of it's interesting A lot of IT people are getting into design thinking and design thinking has gone from being an artisan process to a manufacturing process It's now very linear and very expert based Mary Boo and I are currently writing an article attacking it from a design from a complexity perspective There's no point in having one expert go out and interview people for the reasons. I've just explained They'll only see what they expect to see We focused on making people their own ethnographers mass scalable ethnography So that's kind of like a key principle engage people directly. So some basic principles First of all, the way we make sense of the world is not through grand stories told in workshops there. We're performing Yeah facilitators bias workshops within 15 minutes, by the way Yeah, after 15 minutes the facilitator starts to determine the outcome regardless of people's own views again This is basic one-on-one science guys. All right, learn to live with it. You may not like it but live with it So basically the day-to-day observations the water cooler points that the minor irritations when you first log on Those are the things that inform people not grand stories. It's micro observations micro narratives that matter Secondly text is very limited It's about five percent at best of what somebody knows can be written down That hits big data badly Yeah, big data. I'm going through the third big data hype in my life You know, it's kind of like one algorithm to rule them all in in the darkness bind them Yeah, and everybody believes giving enough money They can write the algorithm which will solve the problem to life the universe and everything and there's no point in doing that We already know it's 42. So let's move on. All right. You either get that one or you don't all right So basically all right mean text is very very limited We know more than we can say we say more than we can write down Yeah, so from our point of view we're as much interested in capturing pictures Drawings and voice as we are in capturing text because that's the day-to-day meaning that we need We also need human beings to interpret it because the machine cannot interpret even text the way the human being who created the text interprets it Because actually the meat the text is just part of what the human being knows. They need to add layers to it Right, so they don't get me wrong. There's huge value in big data. It's just over-hyped Yeah, yeah, people is it's just crazy. I mean with the Internet of things We can do a huge amount of stuff, but we still need human interpretation The work we're doing on diabetes management and obesity management at the moment involves human beings interpreting their experience It doesn't involve machines doing it, but because we've got machines. We can actually scale much faster. That's called augmentation not replacement Thirdly human language if you don't know it art cave painting came before language in human evolution One of the unique things about human language compared with eight languages or Cretacean languages is it evolved from abstractions not from naming things So actually we're far more comfortable with pictures Lean concept. Yeah, or with cartoons or with metaphors That's how we explain things So if you actually want people to give you meaning you need to work at a different at an abstract level not at a concrete level People are far more comfortable dealing with abstractions So for example with water engineers at the moment, we're getting them to report micro anomalies Take a picture and then place it on a triangle Between it smells wrong. It tastes wrong. It feels wrong Now they love that because it's ambiguous and they're dealing with ambiguous data So we're actually increased the amount of errors reported by over 40 Because we've allowed ambiguity of reporting something we're about to move sideways into soft intent software testing If you want to detect weak signals early signs of failure You need to recognize that the whole process is deeply ambiguous and you need to allow people to interpret it in an ambiguous way So to give an example on this. This is from work. We're currently doing on Culture mapping and this is actually by going to everybody in the organization and Saying give an example of a decision made recently which affected you personally which summarizes what it's like to work for this company Yeah, not some pseudo psychological thing Culture is revealed through decisions as I perceive them and then people interpret those decisions onto a series of triangles This is one Where their decision was made by people acting intuitively Interpret in a situation logically always based on principles. This shows from the beta test the distribution by division And you can see this company has got a major problem It's really good at the analytical side But it's gonna have a real problem in a crisis because the ability to make decisions intuitively doesn't seem to be present Now and this is a key aspect of the new method of intervention Instead of saying how do I create an XYZ culture? I now click on that model and I say what can I do to create more stories like this and fewer stories like those Now that's a descriptive not evaluative form of intervention And it's exactly the same thing. We're now doing with user stories What could I develop which would create more user stories like this and fewer user stories like that? Yeah, and that's the pre analysis phase of requirements capture by allowing users to continuously record Every micro experience or a sample of users We can then look at statistical clusters in those experiences and sit down as a development team and say How could I create fewer stories like that more stories like that? That allows technology capability to interact with unarticulated user needs By increasing the degree by which we interact or co-evolve Technology capability with unarticulated needs. We radically improve the funnel that fits into traditional design processes I can show you some examples of that in a minute, but basically description is better than evaluation This is actually an example. This is from one of the US's major manufacturing companies This is actually measuring attitudes of the workforce Now actually on the horizontal dimension we have work completion on the vertical dimension. We have rule compliance So you can start to see why this company has got a problem Yeah, in the factory on the left you either get the job done or you follow the rules in The factory on the right you either get a job done or you've given up trying to do either Now the transition mechanism this the traditionally approach Would be to try and move everything to the top right-hand side with a corporate-wide initiative familiar with that We've got a safety problem big program. We're not doing that anymore If you look at the right-hand side, you can see a transition report So the factory manager is sitting down with the people on the factory floor and saying how would we create more stories like that? and fewer stories like these better than that Each factory manager has their own landscape And because those landscapes are coming in from observers continuously capturing material. It's actually changing in real time So each factory manager can now sit down and say how can we create more like this fuel like this? You start to see how we do object-based architecture now in systems design We have multiple real-time representation of a user or consumer space We're constantly slotting new code and new capability into the system We're getting real-time feedback, which is statistically measurable This is a quantitative technique not a qualitative technique. So we can measure vector not velocity And that's actually important vector not velocity Because actually if things are moving in the right direction then actually we've succeeded if we try and define an unachievable endpoint We are always going to fail That's why failure rates are so high because we're dealing with complex systems not order systems in a complex system You can define direction, but not order Which leads me finally onto a key concept called acceptation Dinosaurs feathers evolved you have to be careful on this in the states You have to start off by apologizing if any evolutionary is offended and then saying you have no intention of actually taking them seriously I quite enjoyed doing that one Basically a dinosaur's feathers evolved for warmth or sexual display Then one day a dinosaur with a lot of feathers falls off a tree and it glides and we get birds Yeah, if an evolution is relied on what are called non-linear Exaptive moments. It's not adaptation. It's acceptation. We see the same in technology Arrathian engineer maintain in a radar machine notices a chalked up our melts in their pocket Yet we get microwave ovens. I Remember when I was a coder. I often had sudden insights of something. I've done somewhere else I could move sideways. That's called acceptation So look at this example on my left User stories interpreted by the users into six triads on my right technical capability Cross silos Interpreted into six trials with similar labels polymorphic by technical people. I Now merge them at a metadata level This is actually how you handle cross-silo knowledge cross-trans Disciplinary stuff. We can get metadata at the right level of abstraction This is a whole new approach by the way to legacy database management Is we create human metadata structures and then we use training databases to actually do machine intelligence? By metadata integration, I can integrate radically different databases very very quickly And what I also find then are acceptive moments where clusters of user stories linked with technology capabilities. I Can now look at that and say why but I've also got an evidence base to do a prototype And I can see cases where user stories don't know with technology capabilities or technology capabilities have no link with user stories This is a dynamic real-time continuous feedback system And that leads me on to my final slide when it relates to scaling So a couple of lessons Complexity is not scaled by aggregation A complex system can't be reduced to the sum of its parts The properties of the whole are different from the properties of the part. So by definition, it's not reductionist and it's not aggregative Now this is actually what has upset a few people in safe because I'm not saying it's wrong pragmatically and saying it's wrong a priori Yeah, you actually can't scale in that way if the system is complex You might get some approximations because good people will make it work despite itself, but it's wrong in principle It's not just wrong in practice So fundamentally, don't worry about the triangle. I'll explain it in a minute The basic principle about complexity is you reduce the granularity and you allow what's called recombination and partial copy If you actually look at evolution, it's partial and incomplete copying which produces progress not exact copy So by actually breaking things down allowing small starting with the core unit user stories technology capabilities and Allowing them to recombine in novel ways. That's how you scale Now on the right-hand side you see an example. We've been talking about recently Getting people in strategy to tell stories about things which are keeping them awake at night Getting users to talk about things which frustrate them getting technology to talk about their capabilities And then you get a three-way acceptance So I don't scale by aggregating up. I scale by finding new patterns across the whole domain And that's this kind like basic principle is we need to engage the real sea level not the sea level in it They're already engaged It's just a game You've got to engage the sea level in strategy in marketing in customer relationships And you're not going to do that by it structures You're going to do that by finding novel solutions to real-world problems and by distributing capture So it's pervasive and non-intrusive rather than forcing people into workshops and interview type processes Engage it's scaling by engagement not by aggregation Learn early fail less. I made that point earlier and this point about continuous real-time feedback Now what I'm trying to take here is an approach to scalability based on science Not based on somebody's perception of practice or need to earn consultancy revenue And we need to get a natural science approach back into software development and that's come out the argument here Because actually at its basics complexity is really very simple But it shouldn't be simplistic If you understand the basic principles, it's very easy to adjust and there are three fundamental heuristics about complexity management reduce the granularity Distribute cognition and disintermediate decision makers that means put in decision makers in direct contact with raw material So coders should have direct contact with user stories. It shouldn't be mediated by analysts Executives should have direct contact with technology which might make a difference strategically because they didn't even know it was possible Disintermediation applies at all levels and that's what scalability is about. It's sometimes called the new simplicity The trouble is simple things are very easy to grasp when you've allowed the world to become far too complicated Thank you very much for your time. Are we on there? Okay, so I was waiting you for you to use the word culture and I always thought you weren't gonna use it at some point Which would have been awesome Because it gets used often in our profession so much and like what kind of culture do you need to support agile teams in the corporation? So I'm kind of I would love to hear What culture means in the context you were talking of it from and how People in our profession can practice with it. I have to be very careful here because I encourage my daughter to do anthropology She's now got an MA in cultural anthropology and I get when you get corrected by a 25 year old daughter You start to panic. It's dad. You need to read this urgently, right? I'm using culture in the sense the way we do things around here, which we all understand But you won't understand until you work with us Yeah, and culture cannot be engineered Yeah, you can't define a culture. There is no such thing as an ideal culture. There isn't a recipe for this Yeah, culture is what it is and culture is revealed by day-to-day stories So if you want to shift to culture, you can't say I want to get here You can say well, I don't like those stories. I like these stories. What can I go and do to make a slight difference? Yeah, so you're managing that sort of direction. That's all you can do. That's the most you can do Hey, one other quick question. What's a micro narrative? Okay, it's a fragmented observation Let me give you an example we did one project with an Air Force So we were looking at how culture impacted on staff retention So we pulled in 3,000 stories over a week from officers and spouses of officers It's actually interesting spouses and older children often tell you more about culture than the people themselves Now that was done on the series I can send you the paper if you want on a series of triangles based on cultural anthropology not on organizational design I don't buy that stuff. We go back to anthropology. So things like attitudes justice attitudes to identity From that we got an absolute correlation between effectively a Disposition to leave the Air Force early and recent implementation of lean six Sigma Yeah, we had a multi-choice question about which major initiative is associated with that didn't surprise me Is anybody who puts lean together with six Sigma doesn't understand lean I mean lean is about eliminating the waste that six Sigma creates in the first place, but never mind, all right Six Sigma is BPR with American Bible Belt cultism added on it to a good measure All right, if you've got a black belt, you know, you should be treated as a heretic in my view All right, but then one so basically we get that and chief of the Air Force does not like this news All right, he's really pissed off because that's his pet initiative So we said okay, and remember I'm not evaluating this has come from his own officers It's his officers interpreting their own stories. So he can't challenge it So he's paying attention So he looks at a young officer at the bottom of the table and he says we by which he means you Haven't implemented our by which he means my initiative properly Right, so now he's blaming somebody and we said we don't know look at the story. So this is the micro narrative It was a non-hypothesis question. It's a key. You don't have a hypothesis when you do narrative work So the non-hypothesis question was you're a grandparent in your grandchild says they want to join the Air Force What will you tell them? Yeah first micro narrative one liner. I'll shoot them first and they'll be grateful and He said, oh my god. He said that's some disgruntling officer. So we don't know click on the demographics 30 year warrant officer Now at that in the moment, there's just silence around the table Then he clicked on the second one to paragraph story says why do we have to shit under the trees? Yeah, and it's a story of a new Air Force base, which has just been set up This is actually in the desert, right? And the new six Sigma Pro sorry the six Sigma process means they haven't got a mobile a tree So they're having to dig holes in the ground Rather than use it and they're pretty pissed off. They're a month into deployment Right two paragraphs and I saw the chief of the Air Force get up go to his desk and have a quiet assertive conversation with somebody Which is always scary in military terms The treens got flown in Yeah, almost immediately, but then he came and sat at the table. He said we've taken six Sigma to far Now the point is it was a small fragmented anecdotes not a Consultancy-enabled workshop The process of gathering the meant he couldn't challenge them and it was descriptive not evaluative So by looking at a description he worked out what to do for himself Now that's the radical shift you move from evaluation to description from aggregation and summarization Into disintermediation and raw data and to be honest the culture word is there because people talk about it But it goes away if you do the right things the culture follows Yeah, yeah, they got people have got it the wrong way around You don't change the culture to get people to do the right things you do the right things and then the culture followed Does that make sense? That's me interesting questions. Hope you guys enjoyed the Keynote this morning. I saw a few people kind of send me messages saying this is all bounces for me I would say that's actually a very good thing If it's all bounces that basically means you're motivated enough to go back and do some homework and try and understand Because this is a very important perspective. I want everyone here to take away, right? So I would take that as a compliment saying it's all bounces. It's good Let's go back do some homework and that's gonna help all of us. All right. Thanks, Dave That was a fantastic keynote as expected