Why Do Corporations Still Exist in the Age of AI? (Part 2)

AI makes small teams look clever, but it doesn’t erase the reasons corporations exist.

One thing I learned from eighteen years of being part of the agile community is that common sense is no match for naïve optimism. Whether it was Extreme Programming, Scrum, Holacracy, Sociocracy, Teal, or Team Topologies, the suggestion that we could simply “let people self-organize” and “keep the managers away” was as persistent and unkillable as the Dutch vision of ever winning the FIFA World Cup.

But reality is a bitch, and hope is no substitute for experience.

In my previous post, I gave five reasons why collaboration needs management (economies of scale, transaction costs, the hold-up problem, team production theory, and legal personhood). Here’s the second part of that story, with another five reasons explaining why a self-organized group of autonomous people can never fully replace a corporation.

6. The Resource Loop

If I could afford to buy myself a staff of twelve robots, I’d already done that long ago. Alas, it seems I need to give a few more talks and workshops before I ever get there.

Some things are simply too big (or too scary) for one person to own or to fund. The joint-stock company (made official in England by the Joint Stock Companies Act of 1844) was a legal gadget for piling up more capital than any single investor had, while spreading the risk thin enough that nobody got wiped out when everything went downhill.

Edith Penrose in The Theory of the Growth of the Firm (1959) made the case that a firm is really a bundle of odd, mismatched resources that together produce something none of them could alone. Their unused capacity is what drives it to grow. Penrose claimed that growth is capped by how fast the management can learn, a limit that is now known as the “Penrose effect“.

Birger Wernerfelt and Jay Barney built their Resource-Based View on top of Penrose’s work. They said a firm’s advantage comes from what it owns that rivals can’t easily copy, such as assets, skills, and processes. (We nowadays call that a competitive advantage or a “moat.”) David K. Teece, Gary Pisano, and Amy Shuen stretched this into “dynamic capabilities,” the skill of sensing what’s coming and seizing opportunities by reshuffling resources and reshaping the organization.

In other words, it’s all about how you use what you own. And it’s much easier for a corporation than for a network of individuals to own and manipulate properties.

Here, again, AI pushes in two directions.

On the one hand, frontier AI is brutally capital-intensive. The sector is pouring money into GPU clusters, data centers, and energy-grid connections. The entry fee is so steep that the firms adopting the strongest AI are overwhelmingly the biggest ones. Those early adopters then grow faster in sales and markups, resulting in a self-reinforcing resource loop: corporations need scale to afford the AI, the AI throws off better returns, and the returns fund more scale. That effect is pretty hard for a network of individuals to emulate.

On the other hand, APIs and AI agents offer world-class capabilities to anyone. Traditional moats dry up faster than the Seine in Paris under a heat dome. A solopreneur can rent frontier reasoning, image generation, voice cloning, coding, marketing, and legal drafting for the price of a streaming subscription, which makes the Resource-Based View’s competitive advantage very hard to maintain for traditional firms when most essential resources are just one API call away for anyone with an Internet connection.

Conclusion: The largest firms fare better than ever. All the others have a problem.

7. Tacit Knowledge

If I sold my one-person business now, the buyer would not be able to run it. That’s because nearly everything needed to generate revenue sits in my head, not on Google Drive.

Some knowledge refuses to fit in a contract. Michael Polanyi named it tacit knowledge in 1966:

The stuff you know in your hands and your gut, learned by doing, impossible to email. Firms are systems for moving know-how between people.

Just watch how a new medicine is made. It rests on the accumulated instincts of thousands of researchers, trial managers, and regulatory specialists marinating their expertise for years. You cannot chop that up into freelance gigs and reassemble it with a few peer-to-peer contracts. The knowledge lives within the organization.

The Knowledge-Based View says that firms beat markets at sharing and replicating knowledge and expertise, especially the tacit kind that lives in heads, hands, and habits rather than documents and that people and markets cannot easily price or transfer.

The problem with loose networks is that they are nimble and forgettable. They’re poor at holding onto tacit knowledge, enforcing standards, making relationship-specific investments, and building the routines that make next year’s work better than this year’s. When value depends on “how we do things around here” rather than generic labor you can rent off the market, the network becomes insufficient.

Now that AI can codify and spread expertise fast, it even strengthens the case for organizations that can build proprietary learning loops and real internal knowledge infrastructure. Though it remains a challenge because skills and workflows rot faster than ever because of “accelerated knowledge decay.” Keeping every capability in-house means paying to re-learn everything, all the time. When your internal expertise expires like fresh milk, renting it from the market looks clever again.

However, AI is also the first automation that can distill and replicate tacit-seeming knowledge, but only with access to proprietary data, customer interactions, internal documents, and digitized workflows. AI is quietly mastering and mimicking the expertise of your best people. That’s firm-reinforcing. And when the company owns that data, AI infrastructure becomes the most valuable asset of all, and we’re back to the Resource-Based View (see above) as the main reason for the corporation to exist.

I’m a founder, intrapreneur, and former CIO who helps leaders diagnose and redesign their operating models for the age of AI—informed by plenty of scar tissue. This article offers the same lens I bring to talks, workshops, and coaching, from a single team to a multinational. Want it applied to yours? Let’s talk. And if you’re just here for the maps, th