Building organisations for the age of block-chain, machines and exponentiality

Scalable collaboration based on nature’s recipe

If you are a futurologist, corporate strategist, business developer or innovator, or an actual platform designer, then this text is for you. In this text I will guide you through the basic concepts and design principles of a completely new type of organisation. From simple crowdsourcing-based, through stigmergically coordinated, all the way to our future: the collaboratively owned decentralised autonomous organisations or blockchain-platforms. Owned by everybody, run by everybody. Keywords: ‘platform’, ‘crowdsourcing’, ‘algorithm’, ‘stigmergic’ and ‘distributed’.

• Part I : more organism, less mechanism
• Part II, transforming existing organisations
• Part III: designing scalable collaboration

We live in a society that ‘processes’ things. Throughout the last centuries mankind has been industrialising its world. Presumably without thinking twice we took the 19th century factory model as the universal template for organising about everything. Probably just because of it’s success in manufacturing. Probably also because this type of organisation is like a ‘machine’ and thus gives managers a (false) sense of control. We are now running not only factories as factories, but also governments, banks, schools, prisons, food-production and practically everything else.

‘Design a hierarchy of management functions on top of the primary workflows, remove as much waste, slack or delay as possible and minimize deviations from the norm.’

The managed, hierarchical organisation. For physical production called a factory, for knowledge work called a bureaucracy. ERP, LEAN, 6-Sigma.

Well, no more. Things have changed. Building organisations this way, is setting them up for disruption. Innovation, primarily technological, has dramatically changed society and now the universality of the traditional organisation model needs to be reconsidered. People are organised through global networks. Production is being decentralised with technologies like 3D printing. Collaboration can take place any time, any place and any context thanks to online platforms. People no longer accept current way of working. The factory model no longer suffices.

It all restarted with things like operating systems and appstores, the first universal information age platforms. It evolved through virtual networks like Facebook into ‘physical’ organisations like Uber, Lyft and AirBNB, the first global sharing platforms. And it will not end there: a glimpse of the future is already given by Bitcoin and La’zooz, the first global shared platforms.

Part I : more organism, less mechanism

Most current organisations are based on the insights of the likes of Frederic Taylor, Henry Ford and Toyota. This made perfect sense for the manufacturing of physical things in an age without internet, when collaboration largely took place same time, same place, same context and distribution started at the gates of the factory. Today it makes less sense.
We have been building organisations as highly optimised and consequently very inflexible machines that are, in contrast to their design intent, completely beyond anyone’s control. Scaling them is difficult, reacting to disturbances is nearly impossible and to make things worse they are also very fragile, because when a machine becomes too optimised, every part of it becomes a single-point-of-failure. Paradoxically the industrial model, certainly when applied to knowledge work (called: ‘the bureaucracy’), is not very effective either as many talents and abilities of the ‘production units’ (we are talking mostly ‘people’ here) are not used, standardized away. Efficiency has a price.

Scalable production through crowd-sourcing

The industrial or bureaucratic model is becoming obsolete, fast. Whole categories are already being reshaped by a new breed of organisation that is based on a different model called ‘the platform’. Exponential organisations are taking over. They adapt faster and scale even better. Platforms are fundamentally different from bureaucracies. An important characteristic of platform based organisations is a trick called crowd-sourcing. Crowd-sourcing is leaving production up to a loosely connected ‘swarm’ of ‘anonymous’ individuals. Not human per se. On the internet you can source an algorithm, an AI or a bot just as easy as you can source a human. The bitcoin network crowd-sources its ledger maintenance to computers of the general public and it is completely imaginable, probably intended, that in 10, 20 years companies like Uber will source not only human drivers but self-driving cars too. That part of this new economy where the consumers of a service are the same crowd as its producers, we call the ‘collaborative economy’.

Internet crowd-sourcing gives a platform-based organisation access to a workforce of least a few billion skilled human individuals, and with the rise of the internet of things, AI and machines this number will sooner rather than later increase to 10 billion, 100 billion maybe even a trillion ‘things’. Ranging from highly specialised sensors, attenuators, algorithms to resources like storage and computing power to advanced AI’s, smart devices and bots. Like the autonomous cars of the Uber example above. The ‘Internet of Things’ amplifies any organisation’s platform strategy.

By itself crowd-sourcing is not a process template. Nor is it a means of coordination. It is neither centralized nor decentralized. It is merely a principle of sourcing production capacity. And while the term was coined in an internet context, crowd-sourcing is not even internet specific either. In fact, crowd-sourcing was pioneered by nature, long before the internet was invented or humans even existed. Social insects are masters of crowd-sourcing. Most biological processes have some kind of crowd-sourcing going on, from the cellular level up to entire ecosystems.

Crowd-sourcing is also not limited to a specific type of product or service being produced. The crowd at Stackexchange.com uses the wisdom of the crowd to form a consultancy agency, their product being ‘answers’. The crowd at Wikipedia.org produces ‘knowledge’. The crowd at Github.com builds software. The crowd on the bitcoin network run the a payment network. The crowd of Uber.com does physical transportation.

Crowd-sourcing can be used for large, atomic tasks that are offered directly to the consumers in which case we call it macro-tasking, but just as well for fine-grained production in which the tasks are increments of a larger tasks, which we call micro-tasking.

If the producer crowd is the same crowd as the consumer crowd (think AirBnB or Lyft), we could call it symmetric crowd-sourcing. Otherwise we could call it asymmetric crowd-sourcing.

Crowd-sourcing it not an ‘outward directed’ process either, rather it is ‘undirected’. A crowd is a crowd; it makes no difference whether the individuals of that crowds are also your ‘normal’ employees. Large corporations can easily use their own employees as a crowd. This is called inner-sourcing. As we will see later, inner-sourcing is one of the best ways to start transforming existing bureaucracies or factory-model based organisations into exponential organisations. Once the relative controllable and safe process of inner-sourcing is succeeding it is easy to start sourcing from external crowds too. A business process re-engineering method for this exists: the TeamPark method. I developed it for Sogeti/Cap-gemini between 2006–2008.

The reason platforms-based organisations scale so well, is that they don’t need to employ a workforce. Everything connected to the internet is a potential member of their workforce. AirBNB is probably as big as Hilton on some key figures but has only a fraction of the employees, possible only a 1/1000th. Platforms also don’t need to own any production facilities or machines as every ‘production unit’ that joins the crowd-sourcing brings it own. The coders on Github all use their own favorite development tools. The drivers of Uber all use their own cars. The platform-organisation just connects and coordinates. Scaling is built into the organisation by design.

The other essential part of the new model is the platform itself. The platform is the facility needed to crowdsource and coordinate tasks. It connects the producers, consumers and products (services) to each other. A platform can be physical, like the roads and traffic signs form a platform for commuters, but it can also be virtual, like Github is a platform for software developers.

For modern internetted organisations a platform most likely takes the form of apps and websites build on so called API’s: programmatic layers that exposes the underlying platform logic to the internet. Today you cannot be an exponential organisation without an API. First generation platforms like social websites often did not expose an API explicitly, but existed only as a website, but today API’s are essential.

-The Kanban-board is a platform for collaboration

Compared to managed collaboration, platform-based or ‘unmanaged’ collaboration inverts task assignment: tasks are pulled from the pool, not pushed by some central authority. This decentralises a lot of the resource planning since it is effectively crowd-sourced too: resources plan themselves. Modern project management demonstrates this with a very useful ‘instrument’ called the Kanban-board, which essentially facilitates team scale crowd-sourcing. It is a platform. Team members can get tasks from the board and bring them to their next state in the workflow, placing the back on the board. The advantages are flexibility, scalability and effectiveness since contributors can choose tasks based on their own capabilities and preferences. The main disadvantage is loss of predictability since it could be that some tasks stay on the board longer than desired. The solution to this is quite simple by the way: keep a crowd of your own at hand. Like employees but employ them through the same platform as the outer crowd.

Stigmergy for even more flexibility

Simple macrotasking like Uber and AirBNB employ, limits the scalability and flexibility of crowdsourcing since every participant must be skilled and equipped to fulfil that exact task in its entirety. The solution is to get macro task results by combining simpler — and thus easier to crowdsource — micro tasks. I call this incremental tasking: combining the output of many micro tasks into an end service or product (macrotask). Crowdsourcing simpler tasks creates more potential scale since it is likely that more agents with limited capabilities exist than with advanced capabilities. The conventional way to do this would be through some kind of central control mechanism. The Amazon mechanical turk platform is an example of this.

A more scalable and flexible way is to also crowd-source task coordination, not just task execution. This way coordination scales along with execution automatically. I call this stigmergically coordinated organisation, after the principle that nature applies to organise collaborations. It is the holy grail of exponential organisations and the first step in going ‘beyond Uber’. With decentralised coordination, an organisation in essence becomes exactly what the name already implies: an organism. Every part of the organisation is adapting and integrating to its local environment because it actually is part of that environment. Opposite to an industrial system that tries to force-fit the local environment and parts into its own mechanism.

This type of collaboration is called ‘distributed’ since tasks are distributed among many participants, another perhaps better name would be ‘scalable’. Some of the largest human systems are based on the principles of stigmergy. The internet and the world wide web themselves operate largely on stigmergic principles giving them their fault-tolerance and scalability. Any social site on the web is inherently stigmergic. The Massive multi-player collaborations in MMORPG’s are stigmergic. As is road transport or commuter traffic. When you are driving a car around, you are taking part in a gigantic stigmergic collaboration. Most natural ‘organisations’ are also based on these principles. From the flocks of birds to the schools of fish, which collectively operate as a super-organism, to ants and social insects building complex structures. You’re entire immune-system has all the characteristics of a stigmergic system. Stigmergic collaborations can produce complex results without centralised coordination. No managers, extremely scalable, very robust.

One of the best current time examples of companies built around distributed collaboration is software development on Github. Anyone can join in the development and add, delete or change any number of lines of code, at any time. Even when someone else is doing the same to the same lines of code. The platform coordinates the efforts and makes sure the result is correct.

I called decentralized control the ‘holy grail’ of the exponential organisation. Yet it is not something very complicated or rare, I argue it is even the most used form of collaboration in the universe. Yet despite its commonness very few systems are explicitly designed to be stigmergic. The reason is probably because the design principles have, as far as I know, never been described formally and the subject is as a consequence not taught at business schools. Designing a bureaucracy or factory is taught on every university, designing distributed collaborations is more like an accident.

The magic of decentralised coordination is in the simplicity of it: it consist of only simple parts and rules. The complexity is in the whole, not in any of the parts. Take for instance ants. Ants don’t performs difficult tasks, they can’t. Yet the ant people as a whole can discover and maintain an optimal food distribution network around their nest. All that is needed is that agents exhibit simple, predictable reflexes on a limited set of signals they detect in their environment. The platform acts as the collective state of all collaborating agents. It makes no assumptions on timing or duration of tasks, it just rearranges if it breaks. The whole process is self-organising and very robust. Without the need for central coordination or management.

Decentralising ownership and control, too

The current generation platform-based corporations, like AirBNB, Uber, Github and such, demonstrate a lot of desirable properties. They also demonstrate a very undesirable property: centralized control over their ‘economies’. Since they determine the rules and control the platform, they form a single-point of censorship, control and failure. Which clearly is not desirable when so many people are depending on them. If left unattended the future will give rise to a few very big platforms controlling everything, and everyone. It is desirable to also decentralize ownership and execution. Thanks to bitcoin technology, this too is possible. It is called the decentralized autonomous organisation or robocorp (better names have yet to emerge).

If the primary function of an organisation is sufficiently primitive, like that of Uber, AirBNB, a lending bank, a broker of some sort, a marketplace or a payment system, then the organisation can be automated. And with bitcoin technology this automation can be decentralised such that business execution is never controlled by a single authority, but instead has to follow the majority viewpoint. With the same technology ownership can be democratised too, so that all parts of the organisation, including the crowdsourced workers, also have their fair share in the earnings and value.

The decentralized autonomous organisation is brought into reality by a recent invention called the blockchain. Blockchain technology makes it possible to distribute the execution of the business logic in such a way that there is no single point of control or execution. No single person or entity can willingly or accidentally deviate from the agreed upon rules. Not from the inside, but also not from the outside (think: hackers). The rules are coded as a computer program and run by many computers simultaneously. A majority vote or consensus mechanism is used to determine if outcomes are correct. Of course the code needs to be open source, at least to every participant, so every participant can verify it.

The first organisation to use blockchain technology for the decentralised execution of its business rules was the bitcoin payment system itself, for which the blockchain was invented. In the next 5–10 years we will see these DAO’s decentralise and democratise many sectors like banking, transport, energy distribution, communication, payments, even large parts of governments. A working example already exists: Lazooz.org is an experimental decentralized autonomous organisation, for lack of a better name (might I suggest: blockchain-platforms?). Basically algorithms run by a network of crowdsourced mobile phones. Lazooz.org can in theory be programmed to do anything that Uber can, but in addition Lazooz.org is owned by the same crowd that produces and consumes the services. It uses a fairly simple algorithm to distribute ownership based on participation. La’zooz lives solely as a network of apps on the smart phones of its users, there are no central servers, headquarters or managers. Decentralized production, decentralized ownership, decentralized wealth. A perfect anarchy. I can see a future in which many social utilities, like energy grids, banks, insurance companies, transport even parts of government are run decentralised. A true democracy.

Part II, transforming existing organisations

If you can’t beat them, join them. It is not difficult to see that conventional organisation are at risk being disrupted. It is adapt or die. To transform an existing bureaucratically organised organisation into an exponential is difficult. Not because there are no good methods or strategies for doing so, there are, but because a factory or bureaucracy was never built for change. And because of vested interests at all non-essential levels of the organisation: managers will passively sabotage the transformation because it makes them obsolete. However, the disruption their organisation is facing, will too.

The reasons traditional organisations are disrupted are easy to identify.

First, they are built on fixed workflows which are internally synchronised, like a machine, if you turn the crankshaft, every other part starts moving in sync. This makes them very ill adapted to the outside world which obviously is not synchronised to the internal clock of the organisation. A problem that doesn’t exist in platform-based or exponential organisations because they are by their very nature part of the outside world.

Second, centralised control and production makes organisations hard to scale. Scaling up production not only means buying more machines, extending facilities, hiring more personal and such, it also mean scaling up control, until some super brain is needed to control everything. The brain itself is probably not even the biggest problem, connecting it to all parts is. Then scaling down would mean getting rid of that all again. This problem does not exist in a platform-based organisation since they are based on crowd-sourcing, externalising that problem to the crowd-sourced workers.

Third, bureaucracies only use a small part of all talent and possibilities of their (typically human) workforce. When every employee is fitted into a limited, fixed function profile, a lot of his talents are not used. An exponential organisation does not have this problem, since task are chosen by workers themselves, based on their own preferences and abilities, which they obviously know best.

The starting point of transformation is where crowdsourceable processes overlap with available technology and right-minded, flexible employees. Not all processes are suited to be crowdsourced. Not every employee can handle the transformation or the new way of working. Not every type of collaboration has supporting technology yet.

Transforming processes from managed to unmanaged (crowdsourceable) is conceptually very simple: just take out the directed communication. Make all communication anonymous. However, not every part of the organisation can be transformed, some processes need to be managed. First identify which processes would benefit most from crowd-sourcing. Typically these are the under-performing, or ‘difficult to manage or scale’ processes. Like software development. Outward facing processing are always good candidates. Like customer care and help-desk.
Next identify which part of your workforce can handle the transformation, mentally but also skill-wise. You need the flexible, multi-talented people first.

Finally find which tools are available for crowd-sourcing the identified processes. Start out using standard products, maybe a suite like Lotus Connections (corporate scale social networking), maybe just a Gitlab installation (software development) or a Question2Answer forum (helpdesk and knowledge sharing). Or just a Kanban-board.
The trick is to see your workforce as a crowd and redesign processes so they can be crowd-sourced to this crowd. Take out as much of the directed and synchronous communication as possible (using the right technology). Allow other employees to join the new crowd-sourced processes. See the TeamPark method, part II for a complete guide on this process.

The Teampark method for transforming factory-models into platforms-models

Social salary

Don’t forget that most people work to earn money. Mastery, purpose and autonomy become important only after one is able to earn a living. So with the new way of working, a new way of paying should follow. Like the drivers of Uber and the home-owners of AirBNB are paid for actual tasks they perform, after they performed them, so should your new crowd. I build a payment system for this you could look into, it’s called Mobbr.com and it divides rewards based on actual performance within a crowd. For instance on Github my system can be used to divides pledges and rewards based on an algorithmic assessment of the work delivered. You need to experiment with this, keeping in mind that rewards are never completely honest but also serve the purpose of stimulating participation. In the end you will have a top-down component in the form of normal salaries but also a bottom-up component, or social salary. Based on participation on the crowd-sourced processes. A very good place to start, for large companies, is the software development department. Take for instance a large bank, which will easily have over 2000 coders. Coders are very used to crowd-sourcing, most of them participate on some kind of open-source project and have there hobby projects on Github. Very good tools exists too. Move all or most of your projects to an internal Gitlab or Github installation and stop working in fixed teams. Just start an internal open-source community. Use Mobbr to pay a fair amount of salary based on contributions to this community. Once this works, it is very easy to extend the crowd with external workforce and your software development has entered the exponential era.

Part III: designing distributed collaboration

When ants leave their nest they will typical start wandering about in (seemingly) random patterns. Until they stumble upon something that needs to be brought back to the nest, such as food. They pick it up and bring it back, and while doings so the leave a pheromone trail, like a breadcrumb path for other ants to follow. The next ant that leaves the nest needs only to follow the trail. It too brings back some food and reinforces the trail. The larger the food pile, the stronger and more attractive the trail becomes and more ants follow it. The process automatically scales up limited only by the amount of ants available. Until the pile gets cleared out, then more ants don’t return with food but start wandering about again, starting new trails to food sources elsewhere. The original trail evaporates and eventually no more ants follow it. As easy as it scaled up, it scaled down again too. The result is extremely robust and large scale self-organizing group behavior. Without central management or pre-planned processes. Ants have been building exponential organizations long before man even organized. We can use antelligence to build even better platforms, this will be one of the very few texts I know of that explains the mechanics needed.

We call the principle ‘stigmergy’, coined by a biologist that actually studied ants and there is a few things you need to know beforehand. Stigmergic systems are by nature complex. Even when they are very simple. In systems theory a complex system is a system that is non-deterministic and non-reductionistic, which means that it cannot be predicted and that the whole is not just the sum of the parts. Simply put, distributed collaboration is less predictable than conventional bureaucratic collaboration. Everything has it price. Generally speaking however: the larger the crowd, the more predictable it becomes.

Now lets try to define stigmergy or the resulting distributed collaboration a bit more formal. Distributed collaboration is based on an environment populated with agents that react to signals they find in their vicinity. Like the habitat of the ants, the ants and their pheromone trails. The basic components of stigmergy are:

• Environment or platform
• Crowdsourced agents
• Signals in the environment
• Reactions or reflexes to these signals

Stigmergy in it purest form is based on the total absence of directed communication. This is the hallmark of all social collaboration: no directed communication, only broadcast and anonymous communications. Directed communication has known, pre-identified receivers, where undirected has not. Communication is done instead through signals which are left in a shared environment, for everyone to use. For instance, on a Kanban board we place tasks, these tasks are not assignment yet. There are no names on the task cards, any team member can pick the task. Likewise, when we are commuting to our work and signal we are about to turn left, the signal is meant for anybody in the vicinity, not for a specific person. Or when you collaborate on a social website, you interact with signs and icons on the website, not in real-time to actual persons.

To summarize this: stigmergic or distributed collaboration is based on autonomous agents, anonymous signals, reactions and a platform. Agents react to other agents or signals in the platform. Part of the reaction could be the creation of new signals.

An example: daily traffic and commuting

One of the best examples to illuminate these building blocks with, is our daily traffic and commuting ‘platform’. Let’s take a simplified view.

The environment consists mainly of:

• roads
• sidewalks
• borders
• fixed signals

In its simplest form there is a limited set of agents:

The agents are:

• motorized vehicles
• nonmotorized vehicles
• pedestrians

There is also a limited set of signals which are either displayed by the platform or by the agents themselves:

• the traffic signs
• stop signs
• interactor signs like hand gestures, honking, lights

All that is needed to get an orderly collaboration going, is simple set of reactions to these signals, which we learn when getting our drivers license:

• Keep at safe distance of other agents
• Follow instructions as indicated by the signals (signs)
• Display signals when appropriate

It is a simple scheme but the result is a massive multi-user real-life self-organising transport system that doesn’t need any central control and has no management-hierarchy.

Stigmergy on online platforms, such a social sites

Now that we have defined general stigmergic collaboration, we can easily see how this relates to websites or online platforms. The agents are obviously the users. The signals are the icons, votes, comments and all other pieces of content that are left behind or changed as a result from their interaction. There is no direct communication (remember: this is the hallmark of every social collaboration). Every user only reacts to contents on the site, in his personalized view. The designer of the platform creates the set of signals and reactions such that a meaningful collaboration results.

Types of stigmergy

From the basic scheme many variations can be derived. One of the simpler variations is that in which agents do not use any signals in the environment but only react to other agents. Birds organising into a flock only react to their nearest neighbours. They try to keep an average distance and fly in the same general direction of their nearest neighbours. What emerges is a flock that behaves as a super-organism. Simple rules, complex behaviour. Flocks scale from several tens to millions of individuals easily.

From the basic scheme many variations can be derived. One of the simplest variations is that in which agents do not use any signals in the environment but only react to other agents. Flocking birds only react to their nearest neighbors. They try to keep an average distance and fly in the same general direction of their nearest neighbors. What emerges is a flock that behaves as a super-organism. Simple rules, complex behavior. Flocks scale from several tens to millions of individuals easily.

Yet another is one in which the environment is showing signals based on global or non-local events. For instance, the traffic system of a city could set traffic lights in one part of the city based on events in another part. Uber-like applications can offer heat maps that indicate customer or taxi density in the city, allowed agents to make smarter decisions. Or when directed signals are used, delivered to specific agents.

Yet another is one where signals are not anonymous. An taxi app for instance can allow customers to choose specific drivers. This type ceases to have the benefits of true stigmergy very quickly.

Yet another is one where some signals are not anonymous. An taxi app for instance can allow customers to choose specific drivers. This type ceases to have the benefits of true stigmergy very quickly.

In pure stigmergy agents are state-less. They maintain no history and will react to the same signal in exactly the same way, every time again. Agents become less flexible / interchangeable when they start maintaining history or an internal state. The less state agents maintain and the more localised signals and reaction are, the more scalable the system is. It makes them context-free and thus relocatable. Agents that just react to signals regardless of where or when they receive the signal, and regardless of which signals they already received, can be substituted by other agents of the same type, even when taken from other contexts. The environment, in which all signals are stored, is their shared state and coordination mechanism. In general it holds that the less state an agent has, the more flexible and scalable the resulting collaboration is.

Patrick Savalle is a full-stack innovator. He developed the world’s first method for transforming organisations into ExO’s, called TeamPark (2008). He designed and coded the world’s first payment system for online collaboration, called Mobbr (2011). Co-authored an early and influential paper on bitcoin (2013).

He can be reached on patricksavalle.com and tweets as @patricksavalle