The Rise of the Synthetic Organisation Part 1

In a glass-walled conference room on the forty-second floor of a Manhattan tower, a chief human resources officer I know sat through a presentation last spring that she later described to me as "the most expensive fiction I've ever witnessed." A major consultancy was unveiling the company's AI transformation roadmap: eighteen months of phases, a governance council, a centre of excellence, and change champions on every floor. The slides were beautiful. The timeline was meticulous. And every person in the room knew — though none said so aloud — that it would never happen.

Not because the plan was wrong. Because the organisation was structurally incapable of executing it.

This is the story that repeats itself, with eerie consistency, across the Fortune 500. Companies announce AI strategies. They hire Chief AI Officers. They allocate budgets that would have been unthinkable five years ago. And then the organism rejects the transplant.

MIT's NANDA initiative published a report last year titled The GenAI Divide: State of AI in Business 2025. Based on more than three hundred publicly disclosed AI initiatives, fifty-two organisational interviews, and one hundred and fifty-three senior leader surveys, its central finding was stark: sixty percent of organisations evaluated generative AI tools, but only twenty percent reached pilot stage, and just five percent reached production — delivering what the researchers described as "little to no measurable impact on P&L."

This was not a fringe finding. Bain & Company's 2024 Transformation & Change Survey of over four hundred executives found that eighty-eight percent of business transformations fail to achieve their original ambitions. The broader digital transformation failure rate has hovered around seventy percent for years.

The money being spent is not trivial. Global corporate AI investment reached $252 billion in 2024, with private investment climbing 45% year over year. Enterprise after enterprise is writing cheques that their operating models cannot cash.

Something else is happening, though. Something the failure statistics don't capture. While the incumbents struggle to absorb AI into their existing architectures, a different kind of organisation is emerging — one that was never built on the old architecture at all.

I have come to think of these as synthetic organisations.

The Analogue Organisation's Immune System

Every mature company has developed what I call an organisational immune system. Not a conspiracy. Not malice. Not even, in most cases, incompetence. Just decades of accumulated institutional memory — compliance frameworks, approval chains, job architectures, performance management systems, procurement processes — that collectively recognise foreign bodies and attack them.

Like a biological immune system, it does not distinguish between a virus and a cure. It simply identifies what is unfamiliar and neutralises it.

The immune response manifests in ways that are instantly recognisable to anyone who has spent time inside a large enterprise. Compliance frameworks that made perfect sense in 2010 now block AI deployment in 2026. Performance management systems punish experimentation and reward the optimisation of existing processes. Job architectures fragment any AI implementation across dozens of roles, each with its own approval chain. Procurement processes that take eighteen months to evaluate a platform that will be obsolete in twelve.

An analysis of Fortune 500 annual reports by workforce analytics firm Orgvue found that these companies incurred approximately forty-three billion dollars in severance charges in 2024 alone. Forty percent made workforce cuts that year. Half of those who cut in 2023 cut again in 2024. The organism churns without evolving. Constant restructuring, reorganising, rightsizing — but not transforming. The cuts treat symptoms. The disease is structural.

Here is the uncomfortable truth: most of the rules that now block AI adoption were not unreasonable when created. They accumulated for good reasons over many years. But an organisation optimised for predictability, compliance, and risk mitigation is precisely the wrong organism to absorb a technology that rewards speed, experimentation, and tolerance for error.

You cannot bolt a jet engine onto a horse-drawn carriage.

The Rationality of Inertia

Inertia is not a character flaw. It is a rational response to incentive structures.

Consider a mid-level manager with quarterly targets to hit and a career to advance. Every system around her — compensation, promotion, performance review — rewards optimising what already exists. Taking risks that might disrupt current revenue puts her personally at risk. The upside of a successful AI initiative is diffuse, shared across the organisation, and accrues over the years. The downside of a failed one lands on her desk next quarter. She is not being cowardly. She is being logical.

So is everyone else. That is how you get organisations full of intelligent people producing collectively unintelligent outcomes. The system is not broken. It is working exactly as designed. It was simply designed for a different era.

Now consider the executive layer. The median tenure of an S&P 500 CEO has declined to 4.8 years; the average is around 7 years. The typical AI transformation, done properly, takes three to five years. The arithmetic is brutal: the executive who initiates a genuine transformation is unlikely to be the one who reaps its benefits. But they will bear every quarter of pain along the way — the disruption, the cost, the inevitable setbacks — while their compensation is tied to short-term financial targets that transformation temporarily worsens.

And then there is the middle manager who has spent twenty years building expertise in a domain that AI can now perform in seconds. This person is not merely being asked to adopt new technology. They are being asked to participate in the obsolescence of the knowledge that defines their professional identity. William Bridges wrote decades ago about the difference between change and transition: change is situational, but transition is psychological. The middle manager standing in front of an AI demo is not simply learning a new tool. He is grieving — grieving relationships, status, and a version of his professional self that no longer fits.

No rational incentive system in most large organisations compensates for that grief. Until it does, the immune system will keep winning.

The Synthetic Alternative

While established organisations fight their own antibodies, a different species of company has emerged — one that never developed the immune system in the first place.

The most cited example is Klarna. In 2022, the Swedish fintech employed roughly 5,500 people. By late 2025, that number had fallen below 3,000 through a hiring freeze and natural attrition, rather than mass layoffs. Its revenue over the same period more than doubled, reaching $903 million in a single quarter. AI now handles work that previously required more than 850 customer service agents. Remaining employees saw their average compensation rise by 60%. The company went public in September 2025.

Klarna's CEO, Sebastian Siemiatkowski, has been unusually direct about what this means. "I think there is a massive shift coming to knowledge work. And it's not just in banking. It's in society at large."

What makes Klarna a synthetic organisation is not that it uses AI. Plenty of companies use AI. What makes it synthetic is that its operating model was redesigned around AI as a foundational capability rather than an incremental tool. Klarna did not add AI to its existing customer service process. It rebuilt the process with AI at the centre and defined the role of humans within it.

This distinction — between adding AI to an existing architecture and building an architecture around AI — is the fault line of the next decade.

The market is already pricing this in. When Anthropic launched specialised workflow plugins for its Claude Cowork suite in late January 2026 — targeting legal, finance, sales, and marketing tasks — the reaction across software, financial services, and asset management stocks was swift: approximately two hundred and eighty-five billion dollars in market value was erased within two trading sessions. Thomson Reuters suffered its worst single-day drop on record.

That sell-off deserves more nuance than the headlines suggested. What the market was reacting to was not the death of enterprise software — it was the recognition that AI is moving from tool to workflow partner, and that the platforms best positioned for this future are those already embedded in the flow of work. Companies like ServiceNow — whose entire architecture orchestrates workflows across IT, HR, customer service, and operations — are building AI agents directly into those workflows, turning themselves from systems of record into systems of intelligence. The threat is not to all software. It is software that merely stores and displays information without orchestrating the work around it.

The synthetic organisations are not coming. They are here.

The Individual Divide

This divergence is not occurring only at the organisational level. It is happening between individuals.

Kevin Roose, the New York Times technology columnist, recently observed on the Hard Fork podcast that the gap between early AI adopters and those who remain cautious "is really diverging." Some professionals are now running AI assistants that orchestrate significant portions of their work. Others are still using chatbots essentially as they did in 2023.

The early adopters are not merely more productive. They are developing entirely different workflows, intuitions about what is possible, and expectations about what work should look like. Within organisations, this creates a new fracture: between those who have internalised the new capabilities and those who have not yet begun.

Deloitte's 2026 State of AI in the Enterprise report, based on a survey of more than three thousand executives worldwide, found that companies broadened workforce access to AI by fifty percent in just one year. But only thirty-four percent of organisations are truly reimagining their businesses around AI. The remaining two-thirds are optimising what already exists.

The equivalent of giving rowers slightly better oars and wondering why they have not become a speedboat.

The Race and the Path Forward

This framing — synthetic versus analogue — might suggest the outcome is predetermined. It is not.

Established organisations possess something synthetic competitors lack: revenue, clients, institutional knowledge, regulatory relationships, and decades of domain expertise. The synthetic competitor has speed and an architectural advantage. The incumbent has resources and incumbency. The question is which advantage proves more durable.

History offers ambiguous guidance. Of the thousands of firms that have appeared on the Fortune 500 since the list was first published in 1955, only forty-nine have remained on it continuously for all seventy years. Kodak invented digital photography and buried it. Nokia dismissed the smartphone. Blockbuster passed on Netflix. In each case, intelligent people looked at visible signals and found entirely rational reasons not to act.

But history also offers counterexamples. Microsoft, under Satya Nadella, transformed itself from a company that missed the mobile shift into one that dominates the cloud and is now the largest corporate investor in AI infrastructure. The transformation took a decade and cost billions in short-term margin compression. It required a CEO willing to cannibalise existing revenue streams and a board patient enough to let him do so. That combination is rarer than most leadership teams care to admit.

The race between synthetic and analogue organisations is ultimately a race between two kinds of patience. The synthetic organisation needs market patience — the patience to build revenue before capital runs out. The analogue organisation needs transformation patience — the patience to endure short-term disruption while rebuilding its operating system. Most boards have neither.

But some are beginning to find it. And they are finding it by starting not with the technology, but with the work itself.

Starting With the Work

If I were sitting across from a chief executive, I would not start with a technology question. I would start with a design question: How does work get done here — and how should it?

The reason most AI transformations fail is not that AI doesn't work. It is that the work was never redesigned to accommodate it. Organisations layer AI onto processes that were designed for humans operating without AI, then wonder why adoption stalls. It is the equivalent of putting a word processor on every desk and still routing memos through a typing pool.

This is not a new insight, but it has received its most rigorous treatment from Professor Sharon Parker at Curtin University's Centre for Transformative Work Design. Parker, an Australian Research Council Laureate Fellow and one of the world's most cited researchers on work design, has spent decades building the evidence base for what she calls SMART Work Design — a framework that identifies five dimensions of well-designed work: Stimulating, Mastery, Agency, Relational, and Tolerable Demands.‍

The SMART model, validated across three studies involving more than nineteen hundred professionals and published in Human Resource Management and MIT Sloan Management Review, is not merely a wellbeing initiative — though its impact on burnout and disengagement is substantial. It is a structural diagnostic for how work is organised. And its implications for AI transformation are profound.

Parker and her colleague Gudela Grote argued in a landmark 2020 paper — now cited over 340 times — that work design matters more, not less, in a digital world. Their research showed that when organisations introduce AI, automation, and algorithmic management without redesigning work around these technologies, the consequences are predictably negative: reduced autonomy, degraded skill variety, increased monitoring, and what aviation researchers call the "out of the loop" problem — where humans are nominally supervising systems they no longer understand well enough to correct when they fail.

The SMART framework gives this problem a practical vocabulary. When AI absorbs the routine elements of a role, does the remaining work become more stimulating — richer in problem-solving and cognitive challenge — or does it collapse into passive monitoring? Does the human retain agency over how the work is performed, or has algorithmic management stripped away discretion? Are the relational dimensions of work — collaboration, mentoring, connection to those who benefit from the work — preserved or eroded? Is the pace of human-AI collaboration producing tolerable demands, or has it created an always-on intensity that accelerates burnout?

These are not abstract questions. Parker's research demonstrated that even small work design interventions can produce outsized results. In a study at Perth hospitals, the Centre for Transformative Work Design introduced five-minute interdisciplinary surgical briefings — a simple change targeting the Relational and Mastery dimensions of the SMART model. The result was a 30% increase in team efficiency, corresponding to an estimated $5–$17 million in savings per hospital if fully implemented.

The lesson for AI transformation is direct: the unit of change is not the technology. It is the work.

The organisations that get this right begin with what organisational design firms like Q5 call "organisational health"—the alignment between strategy, structure, process, culture, and capability. Smart organisations are using work design platforms like Beamible to make visible what people spend their time on—not what their job descriptions say—and then identify which bundles of work should be automated, which should be augmented, and which should be protected as distinctly human.

Parker's SMART model provides the evaluative lens that these tools need. It is one thing to identify which tasks can be automated. It is another to understand whether automating them will produce work that is stimulating, offers mastery, preserves agency, maintains relational richness, and keeps demands tolerable. Without that lens, organisations risk creating a workforce that is technically augmented but psychologically diminished — efficient on paper, disengaged in practice.

This is the work before the technology. And it is where the real transformation happens.

The question is not "What can AI do?" The question is "What should the partnership between people, technology platforms, and AI look like — and how do we design the organisation to make that partnership effective?"

That question requires understanding work at a granular level: which tasks create value, which consume it, where human judgment is irreplaceable, and where AI can take the lead. It requires redesigning incentive structures so that the mid-level manager we met earlier is rewarded for experimenting, not punished for disrupting. It requires creating protected spaces where teams can rebuild processes from first principles. And it requires leaders willing to hold their nerve through the messy, uncomfortable, metric-dipping middle of transformation.

Can a mid-level manager adopt AI, accept the risk that it disrupts her current metrics, and be confident that the organisation will reward her for the long-term value she is creating? Can a division head invest in rebuilding a process from scratch, knowing he will face months of lower throughput, without fearing that the next quarterly review will end his tenure? Can an individual contributor raise her hand and say, "My role should not exist in its current form" without being the first person shown the door?

If the answer is no — and in my experience, it usually is — then that is where the work must begin.

This is not a technology problem. It is a business design problem, a work design problem, and an organisational design problem. The technology, after all, is not the hard part.

The hard part is building an organisation capable of absorbing change at the speed change now demands. One where people, platforms, and AI are orchestrated — not in competition, but in partnership.

You don't need a software update. You need a new operating system.

The synthetic organisations are already running one. The question for everyone else is whether they will design one before the market designs it for them.