The AI Squeeze

We don't know who discovered water, but we're pretty sure it wasn't a fish. The most obvious things in the world are invisible to the creatures swimming in them. For all of modern capitalism, the corporation was water. It was so ubiquitous, so seemingly natural, that nobody in the field of economics bothered to ask the one question that should have come first: why does it exist at all?

It took a 21-year-old kid from London to finally ask it. And the answer he found is about to become the most important idea in your organization.

The Question

In 1931, a young commerce student named Ronald Coase was at the London School of Economics listening to a professor explain Adam Smith's invisible hand — the idea that free markets, operating through the price mechanism, coordinate economic activity with extraordinary efficiency. Coase found it revelatory. He also saw a paradox.

If the price mechanism was so good at coordinating everything — if the open market was the most efficient way to allocate resources — then why did we have companies? Why didn't everyone just contract with everyone else, transaction by transaction, the way the theory implied they should? Why did thousands of people show up to the same building every day and take orders from a manager instead of negotiating every task on the open market?

Surprisingly, no one had asked this. Not Adam Smith. Not Alfred Marshall. The firm was just there, a given, an assumption so deeply embedded that questioning it felt almost absurd.

Coase won a traveling scholarship and crossed the Atlantic to find out. He visited factories — Ford, General Motors, industrial operations of every kind — not as a dignitary but as a broke 21-year-old with a notebook, looking at how work was actually organized on the ground.

What he saw was friction. Everywhere, friction. Finding a supplier takes time. Negotiating a contract takes effort. Verifying quality takes resources. Enforcing an agreement when things go wrong takes even more. Every single transaction in the open market carried costs that the elegant models ignored.

But inside the firm? A manager says "do this" and it gets done. No negotiation. No contract. No search costs. The firm replaced thousands of expensive market transactions with a single employment relationship.

Here was the answer. Firms exist because, for a certain range of activities, it is cheaper to coordinate work under a roof than to buy it piece by piece on the open market. The boundary of the firm — what it does internally versus what it contracts out — settles at the point where the cost of internal coordination equals the cost of going to the market. Coase called these transaction costs.

He wrote it up in 1937 as "The Nature of the Firm." The paper was ignored for thirty years — too simple, too observational, not enough equations for a profession chasing mathematical rigor. It wasn't until the 1970s that other economists picked up the thread and built it into a full body of theory. Coase won the Nobel Prize in 1991, at eighty, and used the occasion to scold the profession for decades of what he called "blackboard economics" — models disconnected from how businesses actually work.

Here is why that matters to you: Coase didn't just explain why firms exist. He explained how big they should be. And if you've been paying attention to the last twenty years of business, his framework explains what you've been watching.

The Moving Boundary

If the firm's boundary sits where internal coordination costs equal external transaction costs, then anything that changes those costs moves the boundary. This is not an abstract idea. We've been watching it restructure business for the past forty years — and it provides an explanation for a lot of recent business phenomena.

The internet lowered the cost of finding suppliers, comparing prices, and communicating across distances. Outsourcing boomed. Offshoring followed. Companies that once made everything in-house started contracting out everything, from manufacturing to customer service. The Coasean boundary contracted because external transaction costs dropped.

Enterprise software — ERP systems, CRMs, project management platforms — lowered the cost of internal coordination. Companies could manage more complexity with fewer people.

Globalization, containerized shipping, standardized protocols, digital payments — every one of these was a transaction cost reduction that reshaped which activities happened inside firms and which happened in the market.

Then the smartphone put a camera and a computer, a GPS, and high-speed internet in everybody's hand. That didn't just lower transaction costs — it obliterated them for an entire class of work. Suddenly any individual could find a job, accept it, navigate to it, document it, invoice it, and get paid — all from a device in their pocket. When the organizational infrastructure required to coordinate a workforce became an app, Millions of people became tiny companies.

And here we are: The explosion of outsourcing. The rise of the gig economy — over 70 million Americans now freelancing, more than a third of the workforce, projected to become the majority by 2027. Platforms like Fiverr, Upwork, Uber, and DoorDash didn't just create new business models — they made it trivially cheap to find, hire, and pay someone outside the firm for a single task. That's transaction costs dropping. The average size of startups has been steadily declining for decades. Midsize companies are being hollowed out — fewer of them, struggling harder, squeezed between nimble small firms below and scaled giants above.

Everything we've attributed to "disruption" or "the future of work" or "the gig economy" can be explained in one place: Coase's boundary moving, exactly as his framework predicts, as the costs of going outside the firm keep falling.

All of those changes, significant as they were, came out of the same box. They shaved costs at the margins. They made existing organizational forms leaner. They did not fundamentally alter what the smallest unit of manageable work looked like.

And we're not done.

The Atom Splits

There is a principle buried in Coase's framework that most people miss: every transaction has to take place around a negotiable unit of work. And the smaller the unit, the larger the proportion of the cost that gets eaten by the transaction itself. This is why we buy eggs by the dozen instead of one at a time. It's also why we buy them at the same place we buy milk, sugar, and — these days — T-shirts and garden supplies. Bundling is not a convenience. It is an economic response to transaction costs.

The same principle explains why companies bundle tasks into roles, roles into departments, and departments into divisions. The transaction cost of managing each piece individually was too high, so we packaged them together. That's not a management philosophy — it's economics.

But technology keeps shrinking the viable unit of work. Look what happened with Adobe and Figma. For decades, Adobe's atomic unit of work was the document — the Photoshop file, the Illustrator file, the InDesign layout. You bought a massive application, installed it locally, created a file, saved it, exported it, emailed it, waited for feedback, incorporated changes, and re-exported. Every handoff was a transaction heavy with friction. The whole creative industry organized itself around that unit: the deliverable, the file, the final proof. And for a long time, Adobe owned it.

Figma shrank the unit of work to the component. A single button. A color token. A comment on one element of one screen. Multiple people work on the same design simultaneously, in a browser, with no install, no export, no version management. The transaction cost of creative collaboration dropped to near zero — and with it, the entire organizational structure that Adobe's model rested upon began to dissolve. Figma's market share went from 8% to 57% in three years. That's not a product victory. That's a Coasean boundary shift.

Now AI is doing the same thing — but to knowledge work broadly. It is changing the smallest unit of work that can be independently executed.

In corporate marketing — to take a concrete example — the smallest practical unit of outsourceable work used to be the asset. A blog post. An email campaign. A landing page. You could hand that to a freelancer or an agency and get something back. Below that level — the research, the drafting, the revision, the formatting, the variant testing — the overhead of briefing and reviewing made it impractical to transact. So companies kept all of that bundled inside a team. Inside the firm.

But AI pushes the atomic unit of work down to something far smaller. A single decision. A color change. A byte, literally. Write this headline. Resize this graphic. Analyze this customer segment. Score this lead. Generate three subject line variants and predict which will perform best. Each of these used to be an invisible five minutes embedded in someone's larger workflow. Now each one is an independently executable task at near-zero marginal cost.

This is not automation in the way we've talked about automation for the last twenty years. Automation replaced repetitive, predictable, manual tasks. What AI is doing is different. It is making cognitive micro-tasks executable without human labor. We're talking about white collar work. And it's doing it on both sides of the Coasean boundary simultaneously.

Inside the firm: AI makes internal coordination cheaper. Fewer people are needed to manage information flow, generate reports, draft communications, and handle the connective work that holds organizations together (like writing memos to the remaining people). This means firms can run leaner. Which means that they must.

Outside the firm: AI makes external capability accessible. A single strategist with the right AI stack can produce what used to require a department. The transaction cost of going to market for sophisticated knowledge work has collapsed.

With pressure from both sides, the firm gets squeezed. We've all felt it; we just haven't known where the pressure was coming from.

What This Means for You

If you manage a team, run a division, or lead a company, this is not a technology question. It is an organizational design question. And it is arriving whether you have a strategy for it or not.

Every role in your organization is a bundle of tasks. Some of those tasks are now executable by AI at near-zero cost. Some of them still require human judgment, relationship context, institutional knowledge, brand sensibility, strategic sequencing — the connective tissue that can't be atomized.

The roles on your org chart were designed for a world where the bundling made sense. Where you needed a human to do the whole bundle because the pieces couldn't be separated and transacted independently. That world is ending.

The question is not whether your org chart will change. It's whether you redesign it deliberately or let it degrade into a structure that no longer reflects how work actually gets done.

Change Your Org Chart Right Now

What follows is a practical exercise. You will need your current org chart — even a rough sketch will work — and access to an AI assistant like Claude, ChatGPT, or whatever tool your organization uses.

Upload your org chart and use the following prompt. It will walk the AI through a structured analysis of your organization using the Coasean framework we've been discussing. What you'll get back is not a fantasy. It's a rigorous decomposition of where your firm's boundary actually sits today, and where it's likely to move.

The Kid's Revenge

Ronald Coase died in 2013, at 102, still sharp, still frustrated with economists who wouldn't look at the real world. He never saw GPT-4. He never saw a marketing team of three outperform a department of twenty. He never saw a solo consultant with an AI stack walk into a boardroom and deliver what used to require an agency.

But he gave us the framework to understand it. The question he carried to those factory floors in 1931 — why does this organization exist in this form? — is the most urgent question in business today. The answer is the same as it always was: because the costs of doing it another way were too high.

Those costs are falling through the floor.

Your org chart was designed for a transaction cost environment that no longer exists. The twenty-one-year-old kid from Willesden, limping through Depression-era factories with a notebook, figured out the principle. The rest is your problem.