The Efficiency Trap: What AI Actually Changes About How Organizations Work

In December 2008, I was looking for my first job. I would graduate, become a banker, work in a big shiny building, relocate to New York City, wear an expensive suit, walk down Wall Street and talk about something important on my Blackberry. I had a plan and it was going to be glorious. I didn’t, however, plan for Lehman Brothers to file for Chapter 11 in September. Markets tanked, hiring froze, every lead I had shut the door on me. Nobody could predict what I graduated into, and I had no idea what I was going to do.

In the aftermath of the global financial crisis, I readily absorbed the standard account of the subprime catastrophe. Greedy banks. Reckless borrowers. Captured regulators. The villain changed depending on who was telling the story, but the basis always stayed the same: someone knew better and chose worse. Accountability was possible if only we had more ethical actors and better controls.

In recent years, however, I’ve settled on a more structural albeit less satisfying take. Strip away the emotion and ask yourself what the individual actors in every system, every organization, every transaction optimize for - what would you have optimized for if you were in their shoes? The mortgage originator was optimizing origination volume. The bundler was optimizing the portfolio. The rating agency was optimizing model outputs. The institutional buyer was optimizing yield. Every person and entity was a fractional operator, an executor of a specific step in a vast and immensely complex system. They were each doing exactly what the system explicitly incentivized for their role. Nobody had a full map. There was no mastermind coordinating the system from the center. The crisis was not a moral failure assembled from bad actors. It was an unwanted outcome predictably resulting from complex systems that nobody consciously designed.

We called it greed because we needed a villain. The structure needed no villain. It just needed enough people who couldn’t see the downstream impact of their actions to the larger picture.

This piece is not about the financial crisis. It’s about the structural condition the crisis exposed. One that exists in every large organization, in every industry, right now. The division of labor made our society hyper specialized, delivering unprecedented and desired efficiency at every step of the way. It also made us cognitively fractional.

That fragmentation has a cost that only surfaces when something breaks.

What the Pin Factory Actually Cost

Specialization brought with it a structural trade - agency for efficiency.

Adam Smith’s pin factory is the founding image of modern economics. One worker draws the wire, another straightens it, a third cuts it, a fourth points it. Ten workers producing thousands of pins a day instead of the hundreds it would have otherwise required. Specialization multiplies output by orders of magnitude. The division of labor has been the most successful productivity mechanism in human history. It is also a cognitive fragmentation machine. These two facts have always coexisted, and we accepted the second as a trade-off for the first.

Smith understood what we were trading away:

“In the progress of the division of labour, the employment of the far greater part of those who live by labour, that is, of the great body of the people, comes to be confined to a few very simple operations; frequently to one or two...

...The man whose whole life is spent in performing a few simple operations, of which the effects, too, are perhaps always the same, or very nearly the same, has no occasion to exert his understanding, or to exercise his invention, in finding out expedients for removing difficulties which never occur. He naturally loses, therefore, the habit of such exertion, and generally becomes as stupid and ignorant as it is possible for a human creature to become.”

— Adam Smith, The Wealth of Nations (1776)

The specialist who draws wire for eight hours a day loses something the craftsman who made the whole pin had: a complete model of what he was making, why each step mattered, and how his work connected to the finished object. The specialists gain speed and volume, but lose the map over time.

Societies and organizations first made this trade consciously at a small scale. The early craftsman gave way to the assembly line worker. Then Taylorism turned Smith’s pin factory observation into management doctrine, codifying the separation of thinking from doing. For two centuries, we accepted this as structural load-bearing. Not a problem to solve but a feature to manage.

We promoted generalists into leadership so that someone, somewhere in the organization, could see enough of the whole to make coherent decisions. We hired managers to coordinate the fragments. And when the coordination failed, we hired consultants to bridge the gap. Aside from the polished narrative about strategic expertise, what McKinsey sells, at its core, is cognitive wholeness. The ability to see across the fragments that the organization’s own structure prevents it from seeing. A hundred-billion-dollar industry is institutional proof that the gap never closed.

Outsourcing offloaded non-core competency tasks and created entirely new specialized industries. Software compartmentalized roles and tasks, creating layers of specialization within departments and sub-departments. We now have a mortgage broker who originates the loan and immediately sells it to a securitization desk that bundles it into tranches, rated by an agency paid by the issuer, sold to an institutional investor managing a pension fund. We now have a primary care physician, a specialist, a hospital system, an insurance underwriter, a pharmacy benefit manager, and a group purchasing organization standing between a patient and the drug their doctor prescribed.

Nobody designed this failure mode. It is the output of scale. And it is not just finance or healthcare; it is in every organization that has grown past the point where one person can hold the whole system in their head.

The Visibility/Proximity Inversion

Organizations didn’t fail to recognize the fragmentation problem. They built an entire layer of hierarchy to manage it. The problem is that the solution created its own structural tax. The people with broad visibility are separated from the actual work by layers of people management, and the people proximate to real problems lack the vantage to act meaningfully.

This is a structural condition that every large organization produces and nobody has a name for. I want to give it one - the visibility/proximity inversion.

As organizations scale, two things happen simultaneously and inevitably. The people closest to the actual work lose the vantage to act meaningfully on what they see. They have proximity without visibility of the bigger picture. And the people who gain the structural position to see across the system become separated from the work by layers of people management, budget cycles, and organizational politics. They have visibility without the proximity to diagnose the areas to prioritize or the bandwidth to act.

A supermarket was struggling with underperforming frozen food sales. Management assumed it was a product issue. Meanwhile, floor staff noticed that shoppers rushed through the aisle in winter; it was physically cold and uncomfortable. An easy solution is heated rails along the aisle floor, but the staff had no visibility to the sales issue, and management had no proximity to the observation. Call it the stranded insight problem. The observation lives at the bottom, the authority lives at the top, and they don’t meet in the right form.

This is not a leadership failure. It is not a culture problem. It is a structural outcome of scaling any human organization past a certain complexity threshold.

The consultant is the expensive, episodic patch. Organizations hire consultants precisely because consultants carry none of the people management overhead. A consultant can walk into an organization, spend eight weeks developing genuine cross-system visibility, and deliver a diagnosis that the organization’s own people could have produced if they had the same structural position. But they don’t. The knowledge and muscle memory built during the consulting engagement rarely cascades. The consultant leaves, and six months later the organization is back where it started.

This is not a critique of consulting. It is a description of a structural gap that consulting fills because nothing else has been able to. For two centuries, the visibility/proximity inversion was the price of scale. You could manage it, but you couldn’t eliminate it.

The underlying constraint is cognitive: human beings cannot hold a complete map of a complex system while simultaneously executing complex tasks within it.

That constraint just changed.

What AI Actually Replaces

Every wave of automation before AI automated execution. The steam engine replaced muscle. The spreadsheet replaced calculation. Enterprise software replaced manual record-keeping. In each case, the tool did more of a task humans had already been doing: faster, cheaper, at greater volume. The structure of work stayed the same. Only the speed changed.

AI is different. This is not a difference of degree from prior automation, it is a difference of kind.

AI doesn’t just execute faster. It can synthesize across the whole. It can hold a cognitive map of a system while a human operator executes within a fragment of it. For the first time in the history of the division of labor, the fractional operator can have the map.

Specifically, to this end, there are two things AI is phenomenally good at: navigating extensive amounts of context, and tool calling.

It’s fairly non-controversial to say AI excels at handling extensive amounts of data. It can identify patterns and extract signal at a scale and speed that human analysis simply cannot match. Where volume, variety, and velocity often overwhelms human cognition, AI operates with consistency and precision.

The often overlooked capability is tool calling. AI can synthesize information, recognize patterns, select procedures, and sequence operations. Tasks that previously required human cognition to determine “how” can now be performed without it. Revisit our supermarket example: the floor staff notices the cold aisle. Previously that observation goes nowhere. They have no access to sales data, no authority to commission a facilities change, no way to connect their perception to the business problem. With AI tool calling, the staff member inputs the observation, AI queries the sales system for frozen food performance, pulls energy cost data for heated rail installation, checks maintenance schedules, cross-references similar interventions at comparable stores, and surfaces a recommendation with a cost estimate. Nobody programmed “when floor staff reports cold aisle, query these five systems in this order.” The AI figured out which systems were relevant and in what sequence on its own.

The gap between visibility and proximity is no longer structural. The supermarket floor staff that noticed the cold aisle can have AI cross reference that observation against company data without needing to see or understand all the context for themselves. A mid-level operator in a logistics network can have real-time synthesis across the entire chain they operate within. AI understands context, calls the right tools, pattern matches against stated goals, and delivers advice or action, all without sacrificing any of the previous efficiency. Diagnosis becomes available at the level of proximity, not just at the level of visibility. Action follows observation.

This is not a productivity story. It is an agency restoration story.

The fragmentation between visibility and proximity was structural because human cognition has limits. You cannot process the whole system in your head while executing complex and specialized fragments of it. AI can. And when the fractional operator gains genuine cognitive agency over the system they operate inside, the organization changes in kind, not just in degree.

The impact to the management layer is equally profound. AI changes processing capacity and tool calling, but it does not change problem selection. A human still has to decide which problem is worth solving, what a good outcome looks like, and whether the system’s diagnosis is actually addressing the right question. AI maintains and contextualizes against the map. The human still needs to decide where to go. The core function of management no longer needs to be limited to visibility of the whole while supervising an army of executors. The cognitive capacity can be freed up to focus on a far more difficult task: clear goal alignment. Decentralized agency restored to operators combined with prior scale efficiencies is a recipe for chaos without direction. Someone needs to define the north star. For any leader reading this: you know this is harder than it sounds.

Progress, at any scale, depends on the number of people capable of identifying and acting on real problems. Fragmentation suppresses that number by design. For two hundred years, we accepted that suppression as a permanent condition.

It isn’t anymore.

---

The subprime crisis didn’t happen because the actors were bad or leaders incompetent. It was a structural failure point from years of harnessing efficiency in exchange for complexity. Thousands of fractional operators, each optimizing their fragment, nobody holding the whole map, nobody designing for the moment when the fragments misaligned catastrophically.

AI is a change in underlying conditions. The visibility/proximity inversion that has been the structural price of scale for two centuries is now a design choice rather than a structural necessity. The supermarket staff can have the map. The logistics operator can see the chain. The mid-level manager can synthesize across the system they operate inside without waiting for a consultant to do it episodically.

You can redesign around that, or you can wait until the misalignment becomes obvious.