First Principles · Scale
Companies Die. Cities Don’t.
An animal slows down as it grows, and so does a company. A city speeds up. The difference is the network that feeds each one, and AI is about to change what that network costs.
John Rector ·
Y = Y0 · Nβ
- Organisms, ¾
- A city’s pipes and roads, about 0.85
- A city’s wages, inventions, crime, about 1.15
- Dashed: exactly in step, 1
You need a yes on something small: a customer wants a refund that sits one dollar outside the policy. Five years ago, when the company was eleven people, you would have turned your chair and asked; today it is eleven hundred, so the question becomes a ticket, the ticket waits for a manager, and the manager is in a meeting about the tickets.
The customer will hear back in three to five business days. You know this is absurd, and you also know nobody did anything wrong. The company did not get lazy. It got big, and big things slow down. That sounds like a complaint about management. It is closer to a law of biology.
The principle
In 1932 an animal scientist, Max Kleiber, plotted how much energy animals burn against how much they weigh, and found a clean line. Energy does not rise in step with size. It rises as mass to the three-quarters power. An elephant ten thousand times heavier than a mouse burns about a thousand times the energy, not ten thousand. Pound for pound, the elephant’s cells do about a tenth of the work.
In 1997 the physicist Geoffrey West and the ecologists James Brown and Brian Enquist explained why. A body is fed by a branching network. Arteries split into smaller vessels, and those into smaller ones, until the last branch, the capillary, reaches the cells. The network has to reach the whole body, and its last branch is the same size in a mouse and a whale. Build a network that way and it grows more efficient as it grows. It also grows slower. The big animal’s heart beats slower, its life runs slower, it lives longer, and then it dies.
West’s group went on to measure cities. A city’s pipes, roads and petrol stations scale like an animal’s veins, a little below one: a bigger city needs less pavement per person. But wages, patents and new businesses scale above one. A city twice the size produces more than twice the ideas. So do crime and disease. The reason is that a city is not only fed by a network. It is made of encounters, and the number of possible encounters grows faster than the number of people. In a big city, life speeds up.
Then they measured companies, and companies looked like animals. They grow, they slow, and their risk of dying does not fall as they age. Cities almost never die.
Whatever is fed by a network grows cheaper and slower. Whatever is made of encounters grows richer and faster.
So what is the network that feeds a company? Not the product. It is the circulation: approvals, reports, schedules, budgets, handoffs, the forms that carry a decision from where it is needed to where it is allowed and back again. Each person needs a share of it, and that share is where size turns into slowness. Your refund ticket is a capillary.
The turn
What AI changes is the price of the plumbing. The work that carries a decision is summarising, routing, checking against a policy, scheduling, and reporting upward. That is exactly the work language models and agents now do in seconds for almost nothing. None of that is news.
The part most people miss is what happens to the shape. If the feeding network costs almost nothing per person, the main reason size meant slow is gone. What remains inside a large organization is the other network, the one cities run on: people meeting people. That network scales above one. So the large firm does not simply get leaner. It starts to grow like a city. Faster with size, richer per person, and harder to kill.
There is a price, and it comes with the shape. Cities scale crime, disease and conflict above one too. And West’s arithmetic says growth above one has no resting point. To keep going, a city has to reinvent itself faster and faster, each reinvention arriving sooner than the last. The city-shaped company will be quicker and longer-lived, and it does not get to slow down.
You work for an organization of nine thousand people, and the refund question takes as long as it takes to type: the answer comes back with the policy, the exception and the name of the person who owns it, and it does not occur to you to ask a manager. What you notice instead is the traffic: three people you have never met started work this week that touches yours, and one of them will be your best collaborator of the year, and one of them will be a problem.
What I expect
Size stops predicting slowness.
Before 2030, some of the fastest answers to small internal questions will come from the largest organizations, not only the smallest. People will stop telling startups that their edge is speed.
Firms will describe themselves like cities.
The org chart gives way to a map: districts, open entry, small units with their own owners, shared infrastructure underneath. The language of departments, which is the language of organs, yields to the language of neighbourhoods.
The city-shaped ones will live longer.
Over the next decade, organizations that hand their circulation to agents and keep encounters open will die less often as they grow. Organizations that use AI to cut heads while keeping the old body will keep dying on the old schedule.
City problems move indoors.
Fraud, rumour and conflict will grow faster than headcount in these firms. A new management job will look less like supervising and more like governing: rules of the road, a fair way to settle disputes, something like zoning for who may build what, and where.
Reorganization becomes continuous.
The gap between reinventions shortens. The annual reorg gives way to standing change, and stability becomes something people remember fondly.
Where I sit
I live in Charleston. It was founded in 1670. It has burned more than once, it was shelled during the Civil War, the earthquake of 1886 shook it hard, and Hurricane Hugo came straight through it in 1989. It is still here, and busier than ever. No company I know of has a record like that. I also run a small company, Charleston AI, and a growing share of the circulation in my own work, the publishing, the scheduling, the checking, now runs on routines I set up once and do not look at again. The plumbing is going to agents. I am watching which shape that leaves behind.
The claim ledger
- Already true
- Kleiber’s line has held across species for more than ninety years. Cities scale above one in output and below one in infrastructure. Companies age like organisms. Routing, summarising and checking, the plumbing, is work AI already does well.
- What has to happen
- Organizations have to hand the circulation itself to agents, not bolt agents onto the old hierarchy, and they have to keep entry open, so people and small units can meet without asking permission. A firm that uses AI to tighten its hierarchy keeps the animal’s exponent.
- Where I am probably wrong
- Maybe the plumbing was not the cause. If firms are mortal for some other reason, such as a founding idea that ages or control that has to sit in one place, cheap circulation makes them cheaper without changing their shape. If, by the mid-2030s, large organizations that run their circulation on agents still slow down with size and still die at the old rate, then I was wrong.
Back to Thursday, 4:40. The customer, the dollar, the ticket. The ticket was never about the refund. It was what it costs a body to get bigger. Take that cost away and the question goes where it went when the company was eleven people: across the desk, to whoever can answer it. The difference is that the desk is now the size of a city.
Bodies get old. Cities get busy.
Background
- Max Kleiber, “Body size and metabolism,” Hilgardia 6 (1932).
- Geoffrey B. West, James H. Brown and Brian J. Enquist, “A general model for the origin of allometric scaling laws in biology,” Science 276 (1997).
- Luís M. A. Bettencourt, José Lobo, Dirk Helbing, Christian Kühnert and Geoffrey B. West, “Growth, innovation, scaling, and the pace of life in cities,” Proceedings of the National Academy of Sciences 104 (2007).
- Madeleine I. G. Daepp, Marcus J. Hamilton, Geoffrey B. West and Luís M. A. Bettencourt, “The mortality of companies,” Journal of the Royal Society Interface (2015).
- Geoffrey West, Scale: The Universal Laws of Growth, Innovation, Sustainability, and the Pace of Life in Organisms, Cities, Economies, and Companies (Penguin Press, 2017).
- On this site: The Cost Is in the Pairs, on why a team’s cost is its pairs; The Part That Cannot Be Split, on the work that has to happen in order.