Hi Reader,

We are now in the third week of the winter cohort of our masterclass Progressive Organizational Design.

This week is all about Viisi: a Dutch financial service company that is radically decentralized.

They are also one of the most outspoken examples of radical transparency I’ve ever seen. Financial transparency. Salary transparency. Role and decision transparency. Performance transparency. Meeting transparency. You name it.

But what makes Viisi truly unusual is not only how they share information. It’s also about how they treat transparency as a living system. How they keep it up to date.

At the same time, I’m working with Stuart Crainer from Thinkers50 on an alternative management history timeline. Because, as Theodore Roosevelt once said:

That feels especially relevant now, with AI shifting from curiosity to inevitability.

And this combination made me realize something.

Viisi, and many progressive organizations like them, might be unusually well positioned for what’s coming next.

Not despite being human-centric.

But because of it.

Let me explain.

And brace yourself. It’s a long read.

The AI revolution

That AI will radically change our working lives is no longer a prediction. It has quietly become an assumption. It gets repeated so often that it almost risks sounding ordinary.

But the tone has shifted again in recent weeks.

The hype cycle is accelerating because of the rise of agentic AI. Every week brings a new headline about AI agents replacing jobs, automating entire professions, and pushing us toward a future where humans are simply too slow and too expensive to compete.

The narrative goes something like this:

It’s a compelling story. Human versus machine. And this time, the machine wins.

But what if that’s the wrong story?

What if we’re not heading toward a workforce being replaced by AI, but toward organizations evolving into networks of autonomous actors, some human, some artificial?

What if agentic AI doesn’t hollow out human-centric organizations, but actually make them more viable?

To see that possibility, we need to zoom out.

We need to step beyond the headlines and look at the longer arc of (alternative) management history.

Industrialization and mass production

In the 19th century, steam power and mechanized production reshaped the world at unprecedented speed. Work that for centuries had been rooted in craftsmanship was pulled into factories. Artisan workshops gave way to centralized production floors.

The factory became the dominant workplace model. Along with it came task fragmentation, hierarchy, and wage labor at scale. Work was broken into smaller and smaller pieces.

The worker became a cog in a larger machine. And with that came the loss of craft autonomy.

It was no longer the craftsman who dictated the rhythm of work.

It was the machine.

No wonder this era sparked early fears of technological displacement. The Luddites were not irrational. They were reacting to a system in which technology was visibly stripping away control and dignity from their work.

By the early 20th century, this logic intensified. Fordism introduced the assembly line. Taylorism introduced scientific management.

Standardization, measurement, and optimization promised massive productivity gains.

And they delivered.

The first workplace alternatives

And yet, even at the height of industrial centralization, something else was happening.

While industrialization generated extraordinary output, the lived experience inside many 19th and early 20th-century factories must have been harsh: long hours of repetitive tasks with little autonomy.

Right at the moment machines were centralizing work, a different group of pioneers began experimenting with alternatives.

• Robert Owen introduced more humane conditions at New Lanark.
• Quaker industrialists such as Cadbury and Rowntree experimented with more egalitarian organizational models.
• Rochdale pioneers formalized cooperative principles that would later influence organizations such as Mondragon.

These were not marginal footnotes.

They were early signals that productivity and dignity did not have to be opposites.

By the 1920s and 1930s, this line of thinking gained intellectual momentum.

The Human Relations Movement, through the Hawthorne Studies led by Elton Mayo and the ideas of Mary Parker Follett, began to challenge the narrow focus on efficiency, and started to argue for more human-centric ways of organizing.

The sociotechnical breakthrough

The period from the 1940s to the 1970s brought another technological shift. Mainframe computers and early IT systems entered the workplace. Control could now be automated.

At first glance, this seemed like a continuation of centralization.

But beneath the surface, a crucial insight was beginning to crystallize: human systems and technological systems must be designed together.

The inflection point did not come from corporate boardrooms. It came from underground, in British coal mines in the 1950s.

Researchers at the Tavistock Institute, including Eric Trist and the Emerys, studied why some mining teams outperformed others.

What they found was counterintuitive to many. The most productive teams were not the most tightly controlled. They were the ones organized as autonomous teams of miners who planned their own work, allocated tasks among themselves, and solved problems locally.

These early forms of what many now call self-managing teams demonstrated something profound: autonomy and productivity were not trade-offs.

They could reinforce each other.

At roughly the same time, W. Edwards Deming showed how decentralized problem-solving fueled Toyota’s rise, while Stafford Beer developed management cybernetics, later articulated in his Viable System Model.

Both streams converged on a similar conclusion.

Decentralization is not chaos. When designed well, it is a system of distributed intelligence.

This was the intellectual birth of decentralized design.

The first large-scale decentralization pioneers

Then came the personal computer revolution.

From the 1970s onward, computers moved from centralized mainframes to individual desks. Spreadsheets appeared. Email reshaped communication. Enterprise software began structuring workflows across entire organizations.

Coordination costs dropped. And when coordination becomes cheaper, something important happens. It becomes easier to distribute decision-making without losing coherence.

A handful of pioneers recognized this opportunity:

• Bill and Vieve Gore built W.L. Gore & Associates around a lattice structure, deliberately rejecting traditional hierarchy.
• Dennis Bakke and Roger Sant organized AES Corporation as a honeycomb of autonomous plants.
• Kazuo Inamori developed Amoeba Management, dividing Kyocera and later DDI into small, self-accounting units operating like internal entrepreneurs.

These were not boutique experiments. They were large-scale organizations.

Others used decentralization as a survival strategy to save their organizations from near bankruptcy.

• Ken Iverson flattened Nucor and turned it into one of the most productive steel companies in the world.
• Jan Wallander decentralized Handelsbanken, replacing rigid budgeting with local autonomy, a philosophy later popularized under the label Beyond Budgeting.

What these firms demonstrated was decisive.

Decentralization could scale.

And crucially, they did not use technology to replace humans or tighten bureaucratic control.

They used it to enable autonomy. They used it to allow trusted people closest to the customers to make the most decisions.

Many small-scale decentralization experiments

Alongside these large-scale pioneers, something equally important was happening.

A wide variety of smaller, more radical experiments began flourishing. They differed in structure, shape, and governance, but shared a similar spirit.

From the 1970s to the 1990s:

• Chris Rufer built Morning Star around the CLOU system, where colleagues negotiated commitments directly with one another.
• Eckart Wintzen developed the cell organization at BSO/Origin, dividing the company into small entrepreneurial units.
• Gerard Endenburg formalized Sociocracy, introducing structured participative governance within his family’s firm.

Around the same time,

• Ricardo Semler transformed Semco into what became known as Semco-style,
• Jean-François Zobrist reorganized FAVI around mini-factories.
• Lars Kolind dismantled hierarchy at Oticon with his spaghetti organization.
• Koldo Saratxaga began developing the NER model at Irizar within the Mondragon cooperative network.

Not all of these companies remain decentralized today. Some reverted.

But that is not the point.

The point is what this wave revealed.

It revealed that radically decentralized organizations do not converge on a single blueprint. They take many forms. Different structures. Different rituals. Different governance philosophies.

At first glance, that diversity can look like inconsistency.

But beneath the surface lies a shared logic.

These organizations operated as networks of autonomous people interacting through relatively simple rules. They relied on transparency, peer coordination, and clear boundaries rather than top-down control. They produced coordinated and adaptive behavior with surprisingly limited central authority.

In other words, they behaved less like machines and more like complex adaptive systems.

Decentralization did not produce one optimal model. It produced many viable configurations, because people adapted locally to their specific context. What worked in a steel mill did not look identical to what worked in a tomato-processing company or a consultancy.

And that variation was not a flaw.

It was evidence of complexity.

The goal of human-centric organizing, then, is not to search for the perfect model and replicate it everywhere.

It is to design conditions under which multiple models can emerge, adapt, and evolve.

The internet wave: communities and markets

Then came the internet.

And with it came smartphones, social media platforms, SaaS tools, blockchain infrastructures, and entirely new ways of coordinating at scale.

As before, the technology opened two paths.

On one path, the internet accelerated globalization, outsourcing, and platform capitalism. It enabled the gig economy, data-driven management, and algorithmic control.

Work could now be tracked, measured, and optimized in real time. The digital layer became a new mechanism of surveillance.

But on the other path, something very different emerged.

Large-scale open-source collaboration became possible. Thousands of strangers could coordinate directly through a shared platform.

For example:

• Linus Torvalds build Linux into a global operating system built through distributed contribution.
• Wikipedia coordinated millions of edits without traditional hierarchy or managerial oversight.

For the first time, radical decentralization was not confined to small groups or isolated firms building their decentralized systems carefully over many years.

It could now operate at a massive scale, and could grow rapidly.

​Buurtzorg may be the clearest organizational example of this shift.

Founded in the internet era by Jos de Blok, he used digital infrastructure to scale rapidly to thousands of small, self-managing nursing teams.

These teams were structured into Wintzen-like cells, but now connected through an online platform that allowed transparency, coordination, and shared learning.

Scaling a radically decentralized organization to more than 10,000 professionals in just a few years became possible.

It was technology that made direct peer-to-peer coordination feasible at scale.

But again, the technology did not replace people. It connected them. It transformed isolated teams into a living community of autonomous teams.

More recently came blockchain and related infrastructures, allowing some companies to push their decentralized models even further.

In certain contexts, internal transaction costs dropped close to zero. Suddenly, internal markets (as first pioneered with Amoeba Management) became technically feasible in ways they had not been before:

• Zhang Ruimin leveraged these waves to restructure Haier into thousands of micro-enterprises transacting internally.
• Kazuma Sekiya introduced the P-Will system at Disco Corporation to develop pricing systems that allocated resources through internal market mechanisms rather than managerial fiat.

Both examples showed that technology made their internal markets more viable not by replacing humans, but by turning them into entrepreneurs within a larger decentralized ecosystem.

What emerged in this wave was something new: networks of small autonomous teams enabled by digital infrastructure.

Some organized as digital communities built on shared purpose.

Others organized as internal markets built on transactions and contracts.

Recurring patterns

If you step back and look across these two centuries of management history, a pattern becomes visible.

Every major technological wave opened two paths.

One path used technology to centralize power, tighten oversight, and increase surveillance. It amplified top-down authority and made control more precise.

The other path used the very same technologies to distribute intelligence, increase autonomy, and strengthen local problem-solving. It reduced dependence on hierarchy and made coordination possible without constant supervision.

The history of technology at work is therefore not simply a story of machines replacing humans.

Workplace pioneers have repeatedly shown that new technologies can decentralize the workplace, increase agency, build community, and design more human-centered organizations.

But history also shows that this outcome is never automatic.

It depends on how we choose to organize around the technology.

That same tension is now unfolding again.

Now: AI as your teammate (or boss?)

We stand at another threshold.

Large Language Models. Generative AI. Autonomous agents.

This time the shift feels different. Previous waves connected people more efficiently. They accelerated communication, lowered coordination costs, and enabled new forms of collaboration.

AI does something more radical. It does not merely connect humans. It participates. AI agents are becoming active actors inside organizations.

For the first time, the question is not simply how technology connects people.

The question is whether the AI agent becomes your new teammate, or your new boss.

The dominant narrative assumes the latter. It repeats the familiar drama of humans versus machines, only louder this time. It suggests that artificial intelligence will replace workers, dissolve teams, and render entire professions obsolete.

But history suggests another possibility. The path of more distributed intelligence and increased autonomy.

Radically decentralized organizations are already structured as networks of autonomous teams. Teams that operate within clear boundaries. Where coordination happens through transparency, shared data, and peer accountability rather than rigid command chains.

From that perspective, what is emerging now is not the replacement of the network.

It is the addition of a new type of node.

Not just human teams interacting with human teams. But human teams interacting with digital agents.

Seen this way, the next evolutionary step is not a rupture. It is a continuation. Organizations become networks of autonomous actors, some biological, some digital, all operating within shared constraints.

In fact, we are already seeing early versions of this in several of our Rebel Cell companies, particularly those with strong digital roots.

The paradox

And here lies the paradox.

The organizations best positioned for this shift are not the most centralized, tightly controlled hierarchies.

It are the ones that already operate as modular networks. The ones that treat people as adults. The ones that decentralize authority, make decision rights explicit, and embrace radical transparency.

Because integrating AI into a rigid hierarchy is hard.

Hierarchies depend on positional authority and informal managerial coordination. They run on meetings, escalation paths, unwritten power structures, and knowledge that sits inside a few experienced heads.

Much of the system works because certain people “just know” how things function.

AI does not thrive in fog.

Integrating agentic AI into a modular decentralized network is far more natural.

In a decentralized system like the one at Viisi, roles are explicit. Boundaries are codified. Responsibilities are transparent. Units coordinate through defined interfaces rather than personal power.

Over the last decade, many radically decentralized companies already adopted digital coordination and governance tools to make these structures visible and dynamic.

Digital tools such as GlassFrog, Peerdom, or Talkspirit allowed organizations to map roles, circles, accountabilities, and decision rights in real time. They externalized what used to live inside managers’ heads.

In these systems, you can ask who is responsible for a product, which roles are working on a topic, or what authority a circle holds, and the system provides a clear answer.

Without fully realizing it, many of these progressive companies have already made their organizational design machine-readable. They have structured their governance in ways that a language model can interpret, reason about, and interact with.

Originally, this transparency served humans. It helped new recruits find their way through the network quickly. It allowed autonomous professionals to coordinate without relying on middle managers as bottlenecks. It made peer-to-peer collaboration possible at scale.

Now it has an unexpected side effect.

It makes the introduction of autonomous AI agents far easier.

This is why I think that the most human-centric organizations may turn out to be the most AI-ready.

An important nuance

Before I romanticize this future, however, there is an important distinction to make.

Human teams and AI agents may both fit naturally into radically decentralized networks. But they will not be governed in exactly the same way.

Radical decentralization has never been a free-for-all, despite what critics assume. The most progressive organizations in our Rebel Cell network operate within clear and sometimes strict boundaries.

Autonomy always exists within these constraints.

Humans in decentralized systems operate within two layers of boundaries. There are explicit boundaries, such as contracts, metrics, and defined accountabilities. And there are implicit boundaries, such as peer accountability, reputation, social norms, and the desire for belonging.

Much of the real regulation in decentralized systems happens informally. Colleagues challenge weak decisions. Peer feedback corrects misalignment. Reputation functions as a form of currency. Social belonging subtly shapes behavior.

These invisible guardrails are deeply human and just as powerful.

AI agents, however, are unlikely to be sensitive to reputation dynamics or emotional cues in the same way. They do not feel embarrassment. They do not seek belonging. They do not fear losing face.

Which means they will require more explicit boundaries. More clearly codified constraints. More formalized guardrails.

And designing those formalized guardrails, without suffocating autonomy, may become the biggest management challenge of the coming decade.

Cheers,

Follow us on:

​​