Thinking Companies

2026-05-09

We are used to thinking of companies as groups of people.

That is true, but incomplete. A company is also a system for turning information into action.

A customer says something. A team notices something. A metric changes. A competitor moves. A founder believes something about the future. The company has to absorb all of that and decide what to do.

Even a small company contains more information than any one person can fully understand. The product has one version of reality. Sales has another. Support has another. Finance has another. The founders have another. The customers have the most important version, but often communicate it indirectly, through behavior, complaints, churn, silence, or money.

The CEO’s job is partly to compress this mess.

What matters?

What changed?

Where are we wrong?

What should we do next?

This compression is incredibly lossy. A lot gets dropped. A lot gets distorted. A lot never reaches the people who need it.

AI will make this better.

Not all at once. Not magically. But over time, companies will become much more able to understand themselves.

Every customer conversation can become part of the company’s memory. Every project can be connected to the strategy it is supposed to serve. Every team can understand how its work affects the whole. Every important assumption can be tracked. Every recurring failure can become visible.

The company will start to have something like an internal model.

This makes small teams much more powerful. A ten-person company with excellent internal computation may be able to operate with the clarity of a much larger company. A large company with poor internal computation may move like a giant with damaged senses.

It reduces noise and increases signal.

The founder still has to decide what matters. The manager still has to build trust. The team still has to care. Humans still have to take responsibility for taste, courage, and judgment.

But the work changes. Less time collecting updates. More time making decisions and creating meaning.

The best companies will be the ones with the cleanest loops between reality and action.

They will listen faster. Learn faster. Decide faster. Correct faster.

A badly organized company with AI will still be badly organized. Maybe worse. AI cant save a company that cannot tell itself the truth.

But a company that is honest, clear, and well-designed may become something new: a system that can see itself, reason about itself, and improve itself. A thinking company.

That seems like a very big deal.