A woman’s mother has a fall. What begins as a medical event becomes, within a week, something considerably larger. There are appointments to arrange and a pharmacy to visit twice a month. Her own working hours change, which changes her income, which changes what she can comfortably cover this month. Transportation turns into a recurring problem rather than a one-time one. Her employer needs to know something about the situation, though not everything. Her insurer needs a different set of facts. The hospital needs a third. Meanwhile her son still has practice on Tuesdays and a school fee coming due.
Every one of these institutions performs its function competently. The hospital treats, the pharmacy dispenses, the employer adjusts the schedule, the bank moves the money. None of them is doing anything wrong, and none of them sees the situation whole.
She does. Because she is the only one who sees it whole, she becomes the mechanism that holds it together. She explains the same circumstances to five organizations that will never speak with one another. She translates a medical event into a scheduling problem, a scheduling problem into an income problem, an income problem into a series of payment decisions, and each of those decisions into consequences somewhere else. None of this work appears in anyone’s accounts. It has no name, no line item, and no price.
I have come to believe that this unnamed work matters far more than it appears to, and that it explains something about institutional structure I spent years describing without fully understanding.
The specialization bargain and what it cost
Most of the institutions around us were built to be very good at one thing. Banks handle money. Hospitals handle illness. Employers handle work. Transit systems move people. Government agencies administer particular programs under particular rules. Alfred Chandler documented how thoroughly this specialization reorganized economic life, and it worked extraordinarily well. Dividing complexity into specialized domains is what allowed societies to operate at a scale no earlier form of organization could sustain.
The bargain came with a cost that was never entered anywhere. Specialization creates boundaries, and life does not respect them. When a situation crosses several boundaries at once, somebody has to connect the pieces, and that somebody is almost always the person. People fill out the same forms repeatedly, carry information between organizations that cannot exchange it directly, reconcile conflicting rules, rearrange schedules, make the calls, move the money, and work out what happens when a decision in one place creates a consequence in another.
The institutions perform their functions. The person does the stitching.
Which gives us a fairly compact way to state the problem: when institutions are organized around categories but life is experienced as a system, the person carries the integration burden.
The part I underestimated for a long time is what that burden was doing structurally. It was not merely an inconvenience for the individual. It was the thing making the whole arrangement work. Every gap between institutions was being bridged by unpaid human labor, and because that labor never appeared as a cost to anyone, the gaps never registered as a problem worth solving. The vertical model has been running on a subsidy it does not pay for and cannot see.
Why the structure held so long
I have been writing about ecosystems for roughly fifteen years. When I started the blog in 2010 I called it Blurring the Boundaries, because what seemed most visible then was industries beginning to intersect. Mobility was connecting with energy, cities, insurance, and health. Financial services were moving beyond banks. Health was becoming entangled with food, technology, data, and everyday behavior. The vertical structures were still standing, but the activity around them was becoming noticeably horizontal.
That led me to ecosystems, and to an argument I still hold: value creation increasingly moves beyond traditional industry boundaries as organizations, governments, technologies, and other participants combine to accomplish things none of them could accomplish alone. The idea itself was not new even then. James Moore had described competition in ecological terms in 1993, and a substantial literature followed.
What I explained less well was the timing. I framed the shift mainly through capability, arguing that science, technology, platforms, and convergence were making it possible for separate industries to come together. That explanation holds up, but it only accounts for half of the story. It tells you why more things can connect. It does not tell you why they now have to.
And it left me with a question I kept circling back to. If the direction was visible that early, why has the transition been so slow?
The honest answer is that the old structure could keep functioning because people kept absorbing its complexity. The coordination problem was already there. It was simply hidden inside households, inside the workarounds employees invented for disconnected systems, inside the extra staff organizations hired to bridge functions their technology never connected, and inside the government programs created to bridge other government programs. The system worked because people compensated for it, continuously and invisibly.
There is an obvious objection here, and it is worth meeting directly. For the past thirty years institutions moved burden toward the individual rather than away, and called it convenience. The travel agent disappeared and the itinerary became your problem. The bank teller was replaced by a machine, then by your phone. Checkout became self-checkout, and tax filing became something you do yourself with software. That was offloading, not absorption, and it happened during a period when people still had absorptive capacity to spare.
That is the condition now changing. Institutions externalize work when people remain an inexpensive place to put it, and they absorb it when the burden starts to constrain their own growth. What seems to be happening is that the person is becoming a less sustainable place to put the work, because the volume of boundary-crossing in an ordinary life has grown faster than anyone’s capacity to manage it. The subsidy is thinning. As it thins, the structure it was quietly supporting comes under pressure to change.
Ecosystems as institutional adaptation
Seen from that angle, ecosystems look different to me than they once did. I used to describe them primarily as a new way to create and capture value, which is true and incomplete. They are also a form of institutional adaptation under pressure.
More of the outcomes we care about no longer sit comfortably inside a single institution or industry. Mobility is not simply a transportation problem, wellness is not simply a healthcare problem, and energy now runs through transportation, buildings, manufacturing, national security, and household economics at once. We keep applying industry labels because that is how our institutions are organized, while the underlying reality resists the division.
The mechanism runs in a fairly straightforward sequence. Systemic change increases interdependence, interdependence produces more boundary crossings, boundary crossings generate coordination work, and when institutions cannot absorb that work, people, businesses, and governments absorb it instead. Sustained long enough, that pressure begins to reshape structure.
I would put the claim carefully, because I have overstated it in the past. I do not think horizontal structures are inevitable. I think systemic pressure increasingly favors them, and that the inherited vertical model becomes progressively less able to contain the coordination burden it generates. That formulation leaves room for friction, regulation, failed models, partial ecosystems, and very uneven adoption, none of which would surprise me, and none of which changes the direction.
It also makes me more cautious about treating ecosystems as the destination. Shared infrastructure, interoperability standards, intelligent agents, new governance arrangements, and institutional forms nobody has proposed yet could all end up carrying more of this than the ecosystem models we can currently describe. The deeper transition is from isolated coordination toward systemic coordination. Ecosystems are one expression of it.
Where the burden actually goes
Saying that complexity moves rather than disappears is only useful if you can say where it moves to. Here a distinction I have used for years does real work: the difference between an orchestrator and a modular producer. An orchestrator controls the experience surface, sets the terms of participation, and determines how the pieces come together. Modular producers supply the capabilities underneath. What makes the resulting structure genuinely horizontal is that no orchestrator can build the full capability set its own ecosystem depends on. Apple’s position rests on thousands of capabilities Apple does not own, which is why the ecosystem exists at all. Michael Jacobides and his colleagues describe this as complementarity without full hierarchical control, and it is the feature that separates an ecosystem from a large firm.
Orchestration is therefore bounded by domain rather than global. Nobody orchestrates a city. What seems more likely is a layered arrangement, with a coordination layer close to the person and the household, domain orchestrators beneath it handling money, care, movement, or learning, and specialized capabilities beneath those. Each layer depends on the ones below it and participates in ecosystems it does not control. That layering is what keeps burden absorption from collapsing into a single platform that knows everything about everyone, and it is a consequence of the structure rather than a preference anyone chose.
It also explains why institutions would take this on, which my earlier writing skated past. Absorbing coordination burden is expensive, creates liability, and requires reach an institution does not currently have. It happens anyway because coordination is where switching costs are migrating. Whoever absorbs the coordination holds the relationship, and whoever declines becomes an interchangeable input. Value in modular systems concentrates wherever the binding constraint sits, as Carliss Baldwin has shown, and coordination is becoming one of those positions.
Which surfaces the tension honestly. Burden absorbed is dependency created. The institution that absorbs the most of someone’s coordination holds that person most tightly, and that is precisely what makes the absorption attractive. The safeguard is not the good intentions of the absorbing institution. It is the layered structure itself, together with the ability to move between coordinators. Portability is what separates absorption from capture, and it is the part most likely to require regulation rather than goodwill.
What General Purpose Technologies actually change
General Purpose Technologies rarely produce their largest effects when they are first dropped into existing structures. Electricity initially substituted for older power sources inside factories, and the productivity gains were modest. Paul David’s account of this, building on Warren Devine’s work on the shift from shaft-driven to electrified machinery, showed that the significant gains arrived only once factories were redesigned around what electric motors made possible, freeing layout from the constraints of line shafting. Computing followed a similar arc, beginning with the automation of existing activities and mattering far more once networks, software, and platforms changed how organizations could operate at all.
Artificial intelligence looks likely to follow the same pattern. Most current activity applies it to individual tasks, which is sensible, since that is where the value is immediate and the change is contained. If AI turns out to be the General Purpose Technology many expect, the larger story will be about what gets redesigned once intelligence can operate across activities that previously had to be coordinated by hand.
Coordination has always been the hardest part of the ecosystem model. Organizations hold different data, incentives, rules, technologies, and definitions of value, and connecting them has historically required carefully predefined interfaces and considerable human effort. AI lowers that cost, because it can interpret context, handle variation, and manage exceptions without every interaction being designed in advance.
It also generates coordination work of its own, and I would rather say so than pretend otherwise. Systems that act on your behalf create new obligations: verifying what was done, correcting it when it is wrong, and understanding the basis on which a decision was made. This is the clearest possible illustration of the point that complexity moves rather than vanishes. If a coordination layer close to the person turns out to be an agent acting for her, then some of the burden has been delegated rather than absorbed, and delegation carries its own supervision cost.
New capability has needed new rules
There is a pattern in the history of burden absorption that I missed for a long time, and it matters more than the technology. Institutions that took on consequential work from individuals generally needed two things rather than one. They needed a new capability, and they needed a legitimacy structure around using it: rules defining the institution’s authority, its accountability, and the person’s ability to understand, challenge, or step out of that authority. Where capability ran ahead of those rules, the absorption tended to be resisted, or regulated until it became acceptable.
The mechanisms that emerged were not the same as one another, which is part of what makes the pattern interesting. Physicians took on the interpretive burden of deciding what a symptom means, and that settled into accepted practice as informed consent doctrine developed through the twentieth century, moving medicine away from unilateral physician authority toward disclosure and patient autonomy. Banks took on the burden of judging whether a stranger was creditworthy, which meant inferring a person’s circumstances from records that person never handed over; the Fair Credit Reporting Act of 1970 answered differently, constraining who may obtain a report and for what purpose, and giving people rights of dispute along with protection when a report is used against them. Deposit insurance is different again, and not a matter of permission at all, since what made absorbed liquidity risk tolerable was a backstop sitting underneath the arrangement rather than any consent a depositor gave. Authority, transparency, protection. Different problems, different answers, each bounding a capability that had grown consequential.
The failures follow the same logic. When Target inferred customers’ pregnancies from ordinary purchase patterns, as Charles Duhigg reported in 2012, nothing about the data access was improper and the inference was accurate. What the company lacked was any standing to draw that particular conclusion about a person’s life.
Which locates the boundary more precisely than privacy language usually does. Observing a transaction is ordinary, and correlating patterns within someone’s own accounts is well established. Institutions also draw contextual inferences under existing legal purposes all the time, so inference by itself is not the line. The line sits somewhere narrower: inferring a sensitive life situation from transactional traces, and then using that inferred situation to coordinate actions across the rest of a person’s life. Seeing a pharmacy charge is one thing. Concluding that someone is caring for a parent, and rearranging their financial priorities on the strength of it, is quite another.
The workable answer is that the person declares the situation rather than the institution deducing it. She says once that she is supporting her mother, which opens a bounded scope for a defined purpose and period: what information can be used, which priorities matter, what counts as routine, what needs her approval, and how she ends it. The institution acts inside that mandate instead of reverse-engineering her life from transaction data.
What makes this more than a compliance mechanism is that the repeated explaining is itself a substantial part of the burden. Today she describes her mother’s situation separately to the hospital, the pharmacy, the employer, the insurer, and the bank. Establishing the context once and letting authorized participants coordinate around it is not a workaround for the consent problem; it is burden absorption in its most direct form. It also resolves the dependency question rather than deferring it, because she stops being the coordination layer while remaining the person who sets the terms on which coordination happens. That is a different arrangement from having the work taken off her by an institution that then decides what her situation means.
A different measure of progress
Most ecosystem discussions concentrate on value creation, revenue, participation, data, scale, and competitive position. Those questions are real, and organizations still have to build economic models that work. But there is another question that I think deserves more weight than it gets.
What burden no longer has to sit on the person?
This is not a case for automating every human decision, and it is certainly not a case for removing agency, accountability, judgment, or recourse. It means identifying the work people are forced to perform only because our institutional structures do not line up with the way life is actually lived. If someone no longer has to enter the same information five times, that is progress. If a family does not have to learn the internal workings of several organizations in order to arrange care, that is progress.
The complexity has not disappeared in either case. It has moved, and the question of where it moved to is now the interesting one. It might have moved into a domain orchestrator that absorbed it properly, into a layer of supervision the person did not previously have, or into a dependency she cannot easily exit. Those are very different outcomes, and only the first is unambiguously progress.
The thread
Looking back, the pieces connect more clearly than they did while I was writing them. Blurring the boundaries was the early signal, and the ecosystem work described the structural response it pointed toward, though I was describing that response before I understood the pressure creating it. The convergence work explained why more domains were beginning to interact, and the systemic-change work explains why inherited structures are now under strain. The burden lens shows where that strain lands before institutions adapt, and the General Purpose Technologies now arriving may supply some of the means to move it back into the system.
The industrial era generated enormous progress by dividing complexity among specialized institutions, and that model is not disappearing quickly. But life has become more interconnected than the structures built to serve it, and when that happens the seams between institutions start to matter as much as the institutions themselves. The next period of structural change probably involves recovering some of the integration that specialization pushed outward.
What I would watch is not how connected our institutions become. It is whether the complexity they were created to manage still ends up sitting with the person.
Much of what evolved my ecosystem thinking is captured in my book – Inside the Next Transition
Inside the Next Transition 
Leave a comment