What happens when the cost of coordination shifts
We spent the last piece examining whether the anxiety people are feeling about AI and jobs may be a surface symptom of something operating at a much larger scale, not just which roles survive, but what happens to the organizational structures those roles were built to sustain. And when those structures start losing their foundational advantages, history suggests the consequences rarely stay contained to the workplace.
That claim deserves more than a gesture at history. So let’s actually look at what history says.
Because this isn’t the first time a technology has changed the cost of coordination so fundamentally that the institutions built around the old cost structure stopped making sense. It’s happened several times. And each time, the pattern has been remarkably consistent, consistent enough that it’s worth tracing carefully before trying to apply it to the present moment.
— — —
The Pattern
The through-line across every major technology-driven institutional shift is deceptively simple. When the cost of a specific kind of coordination drops dramatically, the structures built around that coordination being expensive stop being necessary in the same way. Power and capability redistribute. New organizational forms emerge that are better suited to the new cost environment. And the institutions that dominated under the old constraints either adapt, often painfully, or get outcompeted by structures built for the new physics.
The pattern isn’t subtle, but what makes it easy to miss is that each transition looks different on the surface. Different industries, different centuries, different technologies. The common variable isn’t visible until you ask the right question: what specific kind of coordination just got cheaper?
— — —
The Printing Press: The Cost of Coordinating Knowledge
Before Gutenberg, the coordination of legitimate knowledge, who could produce it, validate it, and distribute it, required enormous centralized resources. Copying manuscripts was expensive, slow, and controlled. The institutions that dominated weren’t just gatekeeping information; they were the only entities capable of coordinating knowledge at scale. That capability was inseparable from their authority.
The press didn’t just make books cheaper. It broke the economics of that coordination monopoly. Within decades, the cost of producing and distributing ideas had dropped so dramatically that the institutional justification for centralized knowledge authority began to erode. The Protestant Reformation was the first mass movement to exploit the new information economics, and the first successful challenge to a monopoly that had been structurally unassailable for centuries. Scientific networks formed across borders. Dissenting arguments could circulate faster than centralized authority could suppress them.
The visible disruption was theological and political. The mechanism underneath was a specific coordination cost collapsing. And the institutions built around that cost being high took roughly 150 years to fully restructure into something stable.
— — —
The Railroad and Telegraph: The Cost of Coordinating Commerce Across Distance
Before the railroad and telegraph, the friction of distance meant that markets were local, supply chains were short, and the organizational unit that made sense was small enough to manage without long-distance communication. This wasn’t a failure of ambition, it was a rational response to the actual cost of coordination across geography.
The railroad and telegraph collapsed those costs. Suddenly you could manage inventory across a continent, execute transactions in near-real time, and run a supply chain that spanned thousands of miles. But, and this is the part that usually gets missed, exploiting those new economics required a new kind of organization. The modern corporation, with its management layers, standardized procedures, and professional hierarchy, wasn’t an inevitable organizational form. It was invented because these technologies forced it. Coordinating trains running at speed on single-track lines, across vast distances, with lives at stake, required something that had never existed before: a formal hierarchy of professional managers with clearly defined responsibilities and standardized procedures that everyone followed regardless of local context.
The organizational form that came to dominate the entire 20th century — the large hierarchical firm, was an adaptive response to a specific coordination cost structure. It wasn’t natural. It was engineered. And it spread because the economics rewarded it: the cost of coordination across distance had dropped, but exploiting that drop required scale, and scale required hierarchy.
— — —
The Internet: The Cost of Coordinating Information and Distribution
The internet decentralized access to information, distribution infrastructure, and production tools in ways that broke the gatekeeping function of a long list of previously dominant institutions. Record labels, newspapers, travel agencies, retail distributors, not because those businesses were badly run, but because the coordination advantages that justified their existence had eroded. The cost of reaching an audience, distributing a product, or accessing information had dropped to near zero.
But the internet also contains the most important cautionary note in this whole series of transitions. Because the initial disruption, blogs replacing newspapers, independent artists bypassing labels, small retailers competing with department stores, looked like permanent decentralization. It wasn’t. Within fifteen to twenty years, power had reconsolidated around a handful of platforms that captured more concentrated control over information and commerce than the gatekeepers they replaced. The coordination cost had dropped, but whoever controlled the new infrastructure captured the advantage.
This is the part of the pattern that matters most for what comes next.
— — —
What AI Is Actually Changing
Each of those transitions changed the cost of a specific kind of coordination. The press changed the cost of knowledge distribution. The railroad and telegraph changed the cost of geographic coordination. The internet changed the cost of information access and market reach.
AI changes something different from all three: the cost of cognitive work itself.
Not communication. Not distribution. Not access to information. The actual work of analysis, synthesis, planning, and execution, the cognitive labor that knowledge organizations run on. A recent piece in the Harvard Business Review argued that AI’s biggest economic effect may not come from automating discrete tasks at all, but from reducing the translation costs between people, systems, and data, the coordination overhead that consumes enormous organizational energy and scales badly with size. If that’s right, AI is attacking something closer to the central nervous system of modern organizational life than most of the current discourse acknowledges.
And the organizational data is already showing it. A large-sample study of more than 3,100 public firms by Columbia Business School researchers found that companies flattened their management hierarchies following AI adoption, reducing layers across organizations at a scale that isn’t noise. A parallel study found firms investing in AI shifted toward flatter structures with proportionally fewer mid- and senior-level employees. The org chart is already changing. Not in theory. In the data.
— — —
What the Pattern Predicts
If the historical pattern holds, a drop in cognitive coordination costs this significant should produce a period of genuine disruption to the institutions built around those costs being high, followed eventually by some form of restabilization, probably around new organizational centers that are better suited to the new economics.
The transition period is the part that’s easy to underestimate. The printing press took 150 years to fully restructure European institutional life. The railroad took 50 years of financial panics, monopoly battles, and regulatory fights before the modern corporate form stabilized. The internet’s disruption phase lasted roughly two decades before the platform consolidation set in.
Those transitions were painful for the people and institutions caught in the middle of them. Not because the technology was malicious, but because institutions built for one cost environment don’t gracefully adapt to another. They resist, they restructure slowly and unevenly, and the people whose careers and identities were built around the old structure experience the transition as a threat to something real, because it is.
This is what the job anxiety in the current AI discourse is actually tracking. Not a simple replacement of humans by machines. A transition of the kind that has happened before, operating at the level of organizational structure rather than individual tasks, and therefore broader and slower and more disorienting than the “which jobs survive” framing can capture.
— — —
The Question the Pattern Can’t Answer
But the historical pattern only takes us so far. And being honest about where it stops is more useful than pretending it answers everything.
Every prior coordination cost transition eventually reconsolidated around new institutional centers. The press disrupted the Church’s knowledge monopoly, but states and eventually new institutions recaptured the coordination of information. The railroad disrupted local markets, but the modern corporation captured the coordination of commerce at scale. The internet disrupted gatekeepers, but platforms recaptured distribution.
Each time, the window of genuine disruption was real. And each time, it closed.
The question AI leaves genuinely open, and where the current discourse is almost entirely silent, is whether cognitive productive capacity has properties that make it harder to reconsolidate than prior coordination advantages. The press could be regulated. Railroads could be monopolized. Platforms could be built. All of those required controlling something with a physical location, a chokepoint, a fence you could build around it.
Cognitive capability distributed through hybrid cognition, the recursive loop between a human mind and a system that rewards iteration and associative exploration, doesn’t have the same kind of chokepoint. You can’t legislate how someone thinks. The means of production here lives partly inside people in a way that land and infrastructure do not.
That might mean the window of genuine disruption and redistribution is longer and more structurally durable this time. Or it might mean the reconsolidation happens through different mechanisms than before, credential capture, infrastructure dependency, regulatory control of the models themselves, and the window closes just as it always has.
Or, and this is the genuinely open possibility that nobody can answer yet, this transition might be different enough in kind that the pattern doesn’t fully apply. That the redistribution of cognitive productive capacity is structurally harder to reverse than any prior redistribution of coordination advantage. That the institutions built around controlling cognitive work at scale face not just disruption but something closer to permanent structural erosion.
Those are three very different futures. The historical pattern tells us the transition will be chaotic, uneven, and painful for the people and institutions caught in the middle of it. It tells us the anxiety is tracking something real. It tells us the job debate is aimed at the wrong layer of the system.
What it can’t tell us is which of those futures we’re actually heading toward.
That question is still live. And the fact that almost nobody in the current conversation is asking it, while everyone argues about which jobs survive, might be the most important thing this series has tried to say.
— — —
Sources
1. AI’s biggest economic effect: coordination costs, not task automation
Choudary, S. P. (2026, February). “AI’s Big Payoff Is Coordination, Not Automation.” Harvard Business Review. Argues that AI’s greatest economic impact will come from dramatically lowering the translation costs that keep teams, tools, and data from working together — rather than from automating discrete tasks.
https://hbr.org/2026/02/ais-big-payoff-is-coordination-not-automation
2. AI adoption flattens corporate hierarchies (3,100+ firm study)
Ewens, M., and Giroud, X. (2025). “Corporate Hierarchy.” NBER Working Paper 34162 / Columbia Business School. Large-sample study of more than 3,100 public firms found that companies reduced management layers following AI adoption. Findings are consistent with theoretical predictions about how AI compresses coordination costs within organizational hierarchies.
https://www.nber.org/papers/w34162
3. AI investment shifts firms toward flatter structures with fewer mid/senior employees
Babina, T., Fedyk, A., He, A., and Hodson, J. (2025). “Firm Investments in Artificial Intelligence Technologies and Changes in Workforce Composition.” In Technology, Productivity, and Economic Growth. Firms investing in AI technologies shifted toward flatter organizational structures with proportionally fewer mid- and senior-level employees, consistent with AI reducing the coordination burden previously carried by management layers.
4. Historical transitions: printing press, railroad/telegraph, internet
The historical claims in this piece draw on well-documented scholarship. Key sources: Eisenstein, E. (1980). The Printing Press as an Agent of Change.Cambridge University Press. Dittmar, J. (2011). “Information Technology and Economic Change: The Impact of the Printing Press.” Quarterly Journal of Economics, 126(3), 1133–1172. Chandler, A. D. Jr. (1977). The Visible Hand: The Managerial Revolution in American Business. Harvard University Press. On the railroad creating the modern corporate form, see also Perrow, C. (2002). Organizing America. Princeton University Press.
5. Coordination overhead in large organizations
Cross, R., Rebele, R., and Grant, A. (2016, January–February). “Collaborative Overload.” Harvard Business Review. Time spent by managers and employees in collaborative activities ballooned by 50% or more over the prior two decades, with many spending around 80% of working time in meetings or responding to colleagues.
https://hbr.org/2016/01/collaborative-overload
— — —