AI Wants a New Theory of the Firm

In 1937, economist Ronald Coase asked a great question: why do firms exist?

His answer—for which he won a Nobel Prize 54 years later—was that coordinating efforts is expensive. Markets require constant negotiation—finding suppliers, agreeing on terms, and enforcing contracts. Firms arise when it becomes cheaper to organize activity internally than to rely on the open market.

For nearly a century, that insight has explained why firms exist, and why they look the way they do—layered, hierarchical, and often slow. But what happens when coordination becomes cheap?

Artificial intelligence is a great tool for doing actual work—writing code, generating text, analyzing data, even taking over the job of annoying customers on the phone.

True enough. But AI not only performs tasks, it reduces the cost of organizing them. It routes information, summarizes decisions, flags anomalies, and increasingly connects people and systems without requiring a chain of human intermediaries. Functions that once required meetings, reports, and layers of supervision can now happen continuously and almost invisibly.

Those functions were the work of “middle” management. Hordes of junior execs beavered away, sending orders down and reports up a many-layered hierarchy. It required many layers because it was impractical for any exec to have much more than half a dozen direct reports.

The modern firm was built around the limits of human-to-human communication. Information does not travel cleanly through multilevel hierarchy organizations. It gets delayed, distorted, and filtered. Decisions slow as they move upward. Execution fragments as instructions move downward. And resistance ensues: “Doesn’t corporate know what we are dealing with (way) down here?”

The characteristic scarcity of the system was the time, effort, and frustration of human communication. As AI turns scarce and expensive communication into a new abundance, we may need a new theory of the firm.

When information can be compressed, prioritized, and distributed instantly, the need for multiple layers to manage that flow diminishes. Already we are seeing early signs.

• In software development, tools likeGitHub Copilot—now used by millions of developers—have been shown to materially increase coding speed, allowing small teams to produce systems that once required far larger groups, reducing the burden of constant internal coordination.
• In infrastructure, platforms such as Datadog aggregate signals across complex systems into a single view, automatically surfacing anomalies and allowing fewer operators to manage more moving parts. Now used in thousands of systems and analyzing and learning from billions of system events, Datadog doesn’t just monitor systems. It replaces meetings about systems.
• The same pattern appears even in legacy systems. Modern logistics systems at companies like UPS dynamically route drivers and track packages in real time. The trucks are still brown but the drivers no longer need layers of dispatch and coordination that once required entire teams.

In each case, the payoff is fewer people needed to coordinate more activity.

The result is not that organizations disappear. It is that the cost of holding them together declines. As that cost declines, the optimal shape of the firm will change. Activities that once required large, tightly coordinated organizations may now be handled by smaller, more loosely structured teams. Functions that once justified entire departments begin to collapse into software.

The scale of an organization’s impact begins to detach from its size.

We are told that artificial intelligence has not yet delivered a productivity boom. The aggregate data, so far, is modest. But technology shifts do not transform an entire economy at once.  The changes appear first in pockets—where a constraint has been eased. Today, the loosening constraint is coordination.

Investors seeking to benefit from the AI productivity boom in its early stages should look for companies positioned at the new bottleneck—those building systems that compress coordination costs across entire organizations. Some will be, glossy, cutting edge, and hi-tech. Some, like UPS, will just be “Brown.”

Firms were built to solve the problem of coordination. AI is beginning to solve it in another way, starting where coordination used to be hardest.

At the Gilder Technology Report, we look for opportunities where technology overcomes constraints, and old scarcities are replaced with new abundance.