Why a hundred years of organisational thinking, from scientific management to Kotter’s Dual Operating System, was built on an assumption that AI is now overturning

I want to start this article somewhere unusual, given the series so far.

Not with AI. Not with Kotter. Not even with the Human Intent thesis.

I want to start in 1911.

That year, Frederick Winslow Taylor published The Principles of Scientific Management. It was a slim book by today’s standards. It was also one of the most consequential books on enterprise ever written. In it, Taylor argued that work, all work, could be scientifically decomposed, optimised, and managed. That worker autonomy was the source of inefficiency, and that the role of management was to break tasks into their smallest components, identify the “one best way” to perform each, and train workers to execute accordingly.

Taylor was not alone. In France, Henri Fayol was developing his own theory of administrative management, focused on organisational structure and the functions of management: planning, organising, commanding, coordinating, controlling. His major work was published in 1916. In Germany, Max Weber was articulating bureaucratic theory in parallel, the hierarchical, rule-based organisation as the ideal form for large-scale administration. Together, this trio gave the early twentieth century its foundational theories of organisation.

But Taylor’s work became the symbolic origin point of something larger. It established a premise that would shape every organisational theory of the next hundred years.

The premise was this: organisations exist to optimise human work, and that optimisation is a science.

Everything that came after – Fayol’s structures, Weber’s bureaucracies, the human relations movement, systems thinking, lean, Agile, Team Topologies, Kotter’s Dual Operating System – has been variation, evolution, or reaction within Taylor’s frame.

That century is now ending.

Why This Matters

Before we go further, I want to be clear about what kind of argument this is.

This article is not a takedown of Taylor, Fayol, Weber, or any of their successors. The opposite. The body of work that grew out of 1911 has been one of the great intellectual projects of modern economic life. It produced the modern enterprise, the modern profession of management, the modern global supply chain, and most of the productivity gains that lifted billions out of subsistence work over the twentieth century. It deserves enormous respect.

What I am arguing is something more specific. The premise that all of this thinking shared, that organisations exist to optimise human work, is being overturned by something that none of these thinkers anticipated, because it would have seemed impossible in their time. Execution itself, the thing that all of their theories were ultimately about, is increasingly moving away from humans.

When the underlying premise of an intellectual tradition shifts, the tradition does not become wrong. It becomes incomplete. The question is no longer “how do we optimise human work?” The question becomes “what does an organisation do when execution is no longer primarily human?”

That is a different question. And it is the question that Taylor’s century cannot answer.

A Hundred Years of Refinement

It is worth tracing the arc, briefly, before we examine where it ends.

Taylor’s scientific management focused on the shop floor. Decompose tasks. Time them. Find the optimal method. Train workers. Reward output. The factory became a machine for converting human effort into product, and management became the discipline of optimising that conversion.

Fayol moved up a level. His administrative theory addressed the structure of management itself: the functions, the principles, the chain of command. If Taylor optimised the work, Fayol optimised the supervision of the work.

Weber moved up another level still. His bureaucratic theory described the ideal form of a large organisation: hierarchical, rule-bound, impersonal, merit-based. The bureaucracy was, for Weber, the rational answer to the problem of coordinating thousands of humans to act in concert.

These three founding theories were soon joined by others. The human relations movement, born from the Hawthorne studies of the 1920s and 1930s, recognised that workers were not interchangeable inputs but social beings whose motivation mattered. Peter Drucker introduced the language of “knowledge work” in the 1950s and “management by objectives” shortly after, shifting the focus from manual labour to cognitive contribution. Henry Mintzberg catalogued different organisational forms in the 1970s, machine bureaucracy, professional bureaucracy, divisionalised, simple, adhocracy, recognising that no single structure fits every context.

The 1980s and 1990s brought lean manufacturing, total quality management, business process reengineering. The early 2000s brought Agile, DevOps, and the platform economy. Charles Handy wrote about dissolving organisational boundaries. Frédéric Laloux mapped post-managerial forms. Matthew Skelton and Manuel Pais introduced Team Topologies, with its discipline around cognitive load and flow.

And in 2014, John Kotter published Accelerate. His Dual Operating System, the integration of a management hierarchy with an agile network, represented arguably the most sophisticated synthesis of a century of thinking. Kotter saw the limits of pure hierarchy. He proposed a structure that combined the strengths of bureaucracy with the speed of networked, purpose-driven action. He emphasised that the two systems must function as one organisation, with constant flow of information and shared people moving between modes.

I have a lot of time for Kotter’s work. I agreed with the diagnosis when I first read Accelerate. I still do. Single-mode hierarchies cannot adapt fast enough. Organisations that treat themselves as machines to be optimised rather than living systems that must change will eventually fail.

But here is the thing about Kotter that is easy to miss: as sophisticated and forward-leaning as the Dual Operating System is, it sits squarely within Taylor’s century. Both of his operating systems are built to organise human effort. The hierarchy coordinates human execution. The network mobilises human creativity. The integration mechanism, shared people moving between modes, is fundamentally human. The whole architecture, however inventive, takes for granted what Taylor took for granted in 1911: that organisations exist to optimise human work.

That is not a criticism. It is a recognition. Kotter’s Dual Operating System represents the high point of a tradition. It is the most refined answer that a hundred years of thinking has produced to the question Taylor first posed. It is also the answer beyond which that question itself starts to dissolve.

The Implementation Story

There is one more thing worth noting before we leave the historical arc.

The Dual Operating System has had a mixed implementation track record. The respected organisation design specialist Naomi Stanford has written candidly about her experience implementing dual operating models, describing it as “stuck in Beckett’s cycle of ‘Ever tried. Ever failed. No matter. Try again. Fail again. Fail better.'” Academic reviews have identified persistent structural problems: the hierarchy controls resources and tends to “quietly, yet systematically kill off the network side” over time; volunteer effort competes with assigned responsibilities for finite human attention; and recognition systems do not reward network contributions, so network energy decays.

Even Kotter himself, in interviews, has acknowledged that “in almost every corporate history, network and hierarchy will be found to have co-existed symbiotically for some period of time before the traditional tendency for the hierarchy to dominate eventually took over.”

This is not a flaw in Kotter’s thinking. It is a flaw in the assumption underneath all of Taylor’s century. When both halves of an operating system are powered by humans, they compete for the same finite human attention, time, and political capital. The hierarchy, controlling resources, tends to win. This was a problem long before AI.

It tells us something important. The Dual Operating System was not just a model that worked beautifully and is now being challenged by AI. It was a model that has been struggling in practice for over a decade, even before AI arrived, because the assumption underneath it, that human-powered systems can be coordinated through other human-powered systems, was always strained.

That assumption has now reached its end.

What AI Actually Changes

Agentic AI systems are beginning to remove execution as the primary constraint, not just in software engineering, as we explored in the previous article on Agile, but across entire organisations. Operations, finance, HR, marketing, product – every function is discovering that work it previously coordinated through people and process can increasingly be delegated to machines.

The evidence for this is no longer one consulting firm’s view. It is a striking convergence across the major analyst houses and research bodies.

McKinsey’s State of Organizations 2026 report, drawing on a survey of more than ten thousand senior executives across fifteen countries, identifies AI and agentic AI as one of three “tectonic forces” reshaping how organisations operate. McKinsey’s separate research on the agentic organisation describes how a human team of just two to five people can already supervise an “agent factory” of fifty to one hundred specialised agents running an end-to-end business process.

Deloitte’s Tech Trends 2026 reaches a strikingly similar conclusion. Their framing is direct: “If you just take your existing workflow and try to apply advanced AI to it, you’re going to weaponize inefficiency.” Deloitte refers to AI agents as a “silicon-based workforce” working alongside the human workforce.

Gartner found that evolving the HR operating model accounts for twenty-nine percent of predicted AI productivity gains. Not the tools. Not the training. The structure itself. Gartner also predicts that fifteen percent of daily work decisions will be made autonomously through agentic AI by 2028, and that more than forty percent of agentic AI projects will fail by 2027 because organisations are automating broken processes rather than redesigning operations.

BCG has reached the same conclusion from a different angle: most AI efforts stall because organisations fail to redesign roles, workflows, and governance for human-AI work. Forrester’s research confirms 2026 as the breakthrough year for multi-agent systems.

The pattern across all of these is identical, even though the language differs. McKinsey’s “humans above the loop.” Deloitte’s “silicon-based workforce.” Gartner’s “structure itself.” BCG’s “redesign roles, workflows, and governance.” They are converging on the same insight: the constraint has shifted from coordinating human execution to governing intent across humans and machines.

That is exactly the shift this series has been describing. Govern Intent. Delegate Execution.

What Taylor Could Not See

The reason this is the end of Taylor’s century, rather than just another refinement within it, is that AI undoes the foundational assumption.

Taylor’s premise, that organisations exist to optimise human work, only made sense when humans were the only viable executors. Once you assume humans are doing the work, of course you should optimise how they do it. Of course you should structure organisations to coordinate them effectively. Of course you should develop methodologies to govern that coordination. Every theory in Taylor’s tradition follows logically from that starting premise.

But the premise itself was contingent on the technology of the era. It was true because there was no alternative.

For the first time in history, that is no longer the case.

Agentic systems can interpret intent expressed in natural language. They can decompose complex objectives into actions without human guidance. They can execute those actions, measure their effects, and refine their approach. They can coordinate with other agents. They can operate continuously, adaptively, and at a scale that no human team could match.

This is not a productivity improvement. It is a structural change in what organisations need to coordinate. The human work that Taylor’s tradition was built to optimise is, increasingly, not the bottleneck. It is not even the primary thing being done.

The hierarchy, the bureaucracy, the matrix, the network, the dual operating system, the platform-aligned topology, all of these were sophisticated answers to the problem of coordinating human execution. As that execution moves elsewhere, the elaborate architectures designed to coordinate it become, increasingly, scaffolding around something that no longer needs the scaffolding.

This is what Taylor could not see. It is what Fayol, Weber, Drucker, Mintzberg, Handy, Laloux, Skelton, Pais, and Kotter could not see. Not because they were not brilliant, but because the assumption was so fundamental that it was invisible. Like water to a fish.

What Survives

A hundred years of organisational thinking does not get thrown away. Almost none of it does. The diagnoses survive even as the prescriptions become incomplete.

Taylor’s insight that there are better and worse ways to do things, and that systematic study can reveal them, remains true. It is now applied to AI agent design rather than to manual labour, but the underlying discipline endures.

Fayol’s articulation of management functions, planning, organising, coordinating, controlling, remains useful, even though the things being planned, organised, coordinated, and controlled have changed.

Weber’s insight that large-scale coordination requires rules, roles, and merit-based selection remains the foundation of modern governance, even as the rules and roles increasingly apply to systems rather than just to people.

The human relations movement’s recognition that workers are social beings, not interchangeable inputs, has shaped how we think about culture, motivation, and engagement.

Drucker’s emphasis on knowledge work and management by objectives anticipated much of what intent-driven approaches now formalise.

Mintzberg’s structural variety reminds us that no single organisational form fits every context.

Handy’s foresight about dissolving organisational boundaries has been borne out by the platform economy.

Laloux’s documentation of post-managerial forms has shown that radical decentralisation is possible.

Skelton and Pais’s discipline around cognitive load and flow translates remarkably well into AI-augmented contexts.

And Kotter’s central insight, that organisations need both stability and adaptability, that change cannot be managed through the hierarchy alone, that organisations are living systems, is more relevant now than when he first articulated it.

All of this survives. What ends is the underlying assumption that bound it all together: that the primary thing organisations exist to coordinate is human work.

What Comes Next

Once the underlying assumption shifts, the question changes.

For a hundred years, the question was: how do we organise humans to do work effectively? Every theory in Taylor’s tradition was an answer to that question.

The new question is: how do we organise organisations to govern intent effectively, when much of the execution is delegated to systems?

That is not a question Taylor’s tradition can answer, because the question presupposes a world that Taylor’s tradition could not imagine. Hierarchies, networks, dual systems, topologies, holacracies — all of these are structures for coordinating human work. None of them is, fundamentally, a structure for governing intent across a hybrid human-machine enterprise.

This is what the Human Intent series has been arguing for, article by article. The shift from managing work to governing intent. The death of capital-A Agile and the survival of lowercase-a agile. And now, the end of the underlying intellectual tradition that produced both — and the beginning of something that has yet to be named.

In the next article, we will explore what stops being the unit of organisation when this shift fully arrives. The answer is, perhaps unsurprisingly, work itself. Not as something humans do — that continues — but as the organising principle around which enterprises have been built for a hundred years.

Beyond Taylor’s Century

Kotter saw further than most. So did Fayol, Weber, Drucker, Mintzberg, Handy, Laloux, Skelton, and Pais. They contributed to one of the most consequential intellectual traditions of modern economic life. The world they helped create – the world of structured, professional, scalable enterprise – was one of the great achievements of the twentieth century.

But the assumption that bound their work together is now being undone. Not by a better theory of management, and not by a more sophisticated organisational structure. By a technology that, for the first time in history, can take on much of the execution that those theories were ultimately about.

What replaces Taylor’s century is not yet fully visible. But the shape of it is beginning to emerge.

It is governance, not coordination. Intent, not work. Purpose, not process. Humans, as McKinsey has put it, operating “above the loop” rather than within it.

Govern Intent. Delegate Execution.

That is the shift. And it is the end, with deep respect for everything they built, of Taylor’s century.

This is the third article in the Human Intent series. In the next article, we will examine the deeper consequence of this shift: what happens when work itself stops being the unit around which organisations are built.

Frequently Asked Questions